You are on page 1of 173

U.S.

CHEMICAL SAFETY AND HAZARD INVESTIGATION BOARD

INVESTIGATION REPORT
Final Report

E.I. DUPONT DE NEMOURS & CO., INC.


BELLE, WEST VIRGINIA

METHYL CHLORIDE RELEASE


JANUARY 22, 2010

OLEUM RELEASE
JANUARY 23, 2010

PHOSGENE RELEASE
JANUARY 23, 2010
One Fatality
One Confirmed Exposure
One Possible Exposure

KEY ISSUES:
MECHANICAL INTEGRITY
ALARM MANAGEMENT
OPERATING PROCEDURES
COMPANY EMERGENCY RESPONSE & NOTIFICATION

REPORT NO. 2010-6-I-WV


SEPTEMBER 2011
E.I. DuPont de Nemours & Co. Final Report September 2011

CONTENTS
1.0 INTRODUCTION ......................................................................................................................... 13
1.1 Background .................................................................................................................................... 13
1.2 Investigative Process...................................................................................................................... 14
1.3 E.I. DuPont de Nemours & Co., Inc. ............................................................................................. 16
1.3.1 Company History .............................................................................................................. 16
1.3.2 DuPont Business Areas and Corporate Management ....................................................... 16
1.3.3 Safety at DuPont ............................................................................................................... 17
1.4 DuPont Belle Plant......................................................................................................................... 20

2.0 METHYL CHLORIDE RELEASE (JANUARY 22, 2010) .......................................................... 22


2.1 Background .................................................................................................................................... 22
2.1.1 Methyl Chloride ................................................................................................................ 23
2.2 Incident Description ....................................................................................................................... 24
2.2.1 ECF Sensor Alarm ............................................................................................................ 25
2.2.2 Odor Detection Considerations ......................................................................................... 25
2.2.3 Incident Response ............................................................................................................. 26
2.2.4 Community Impact ........................................................................................................... 29
2.3 Incident Analysis ........................................................................................................................... 29
2.3.1 Mechanical Integrity ......................................................................................................... 29
2.3.2 Design and Maintenance of Rupture Discs ....................................................................... 31
2.3.3 Previous Incidents of Rupture Discs Bursting .................................................................. 32
2.3.4 Management of Change--Technology and Subtle Change ............................................... 34
2.3.5 F3455 Unit Turnaround .................................................................................................... 35
2.3.6 Second-Party Process Safety Management Audit............................................................. 38
2.4 Key Findings .................................................................................................................................. 39
2.5 Root Causes ................................................................................................................................... 39

3.0 OLEUM RELEASE (JANUARY 23, 2010).................................................................................. 40


3.1 Background .................................................................................................................................... 40
3.2 Incident Description ....................................................................................................................... 40
3.2.1 Incident Response ............................................................................................................. 42
3.3 Incident Analysis ........................................................................................................................... 42

ii
E.I. DuPont de Nemours & Co. Final Report September 2011

3.3.1 Reconstructive Analysis ................................................................................................... 42


3.3.2 Pipe Testing and Analysis................................................................................................. 45
3.3.3 Previous Incident Investigation......................................................................................... 46
3.3.4 PM Program Recommendation from 2009 Incident ......................................................... 47
3.3.5 Mechanical Integrity ......................................................................................................... 47
3.3.6 Heat Tracing Design ......................................................................................................... 48
3.4 Key Findings .................................................................................................................................. 48
3.5 Root Causes ................................................................................................................................... 49

4.0 PHOSGENE RELEASE (JANUARY 23, 2010) ........................................................................... 49


4.1 Background .................................................................................................................................... 49
4.1.1 Phosgene ........................................................................................................................... 49
4.1.2 Phosgene Stainless Steel Hose Transfer Operation .......................................................... 50
4.1.3 Phosgene Highly Toxic Material Guardian Committee .................................................... 56
4.2 Incident Description ....................................................................................................................... 57
4.2.1 Community Impact ........................................................................................................... 61
4.3 Incident Analysis ........................................................................................................................... 61
4.3.1 Hose Failure Analysis ....................................................................................................... 61
4.3.2 Effect of Plastic Adhesive Tape ........................................................................................ 62
4.3.3 Hose Degradation Issues ................................................................................................... 64
4.3.4 Hose Change-out Frequency ............................................................................................. 64
4.3.5 SAP Work Process ............................................................................................................ 66
4.3.6 Near-Miss Phosgene Incident ........................................................................................... 67
4.3.7 Mechanical Integrity ......................................................................................................... 68
4.3.8 Flex Hose Materials of Construction ................................................................................ 70
4.3.8 Non-routine Job Planning ................................................................................................. 72
4.4 Process Hazard Analysis ................................................................................................................ 73
4.5 Audits ............................................................................................................................................. 75
4.5.1 Unit Second-Party PSM Audit .......................................................................................... 75
4.5.2 2006 Phosgene Committee Audit ..................................................................................... 79
4.6 Standards and Guidelines............................................................................................................... 80
4.7 Key Findings .................................................................................................................................. 86
4.8 Root Causes ................................................................................................................................... 87

iii
E.I. DuPont de Nemours & Co. Final Report September 2011

5.0 THREE INCIDENTS IN 33 HOURS ............................................................................................ 88


5.0.1 Additional DuPont Incidents............................................................................................. 90
5.1 Management Systems .................................................................................................................... 91
5.1.1 Knowledge Management .................................................................................................. 91
5.1.2 Hierarchy of Controls ....................................................................................................... 93

6.0 REGULATORY ANALYSIS ........................................................................................................ 97


6.1 Occupational Safety and Health Administration (OSHA) ............................................................. 97
6.1.1 Process Safety Management Program .............................................................................. 97
6.1.2 Compressed Gases ............................................................................................................ 99
6.1.3 Inspection History ........................................................................................................... 100
6.2 Environmental Protection Agency ............................................................................................... 101
6.3 State Hazardous Chemical Release Prevention Program ............................................................. 103

7.0 RECOMMENDATIONS ............................................................................................................. 105

REFERENCES ......................................................................................................................................... 112

APPENDIX A: THREE EVENT LOGIC TREE ...................................................................................... 114

APPENDIX B: HISTORICAL AND EVENT TIMELINE ...................................................................... 119

APPENDIX C: SAP PROGRAM ............................................................................................................. 120

APPENDIX D: PHOSGENE RELEASE CALCULATIONS .................................................................. 123

APPENDIX E: HAZARD ANALYSIS FOR PHOSGENE USE AT BELLE ......................................... 130

APPENDIX F: HARD PIPE TO FLEXIBLE HOSE TRANSITION CORRESPONDENCE ................. 130

APPENDIX G: PHA RECOMMENDATION DELAY LETTER ........................................................... 165

iv
E.I. DuPont de Nemours & Co. Final Report September 2011

List of Figures
Figure 1. DuPont Belle, WV, facility on the Kanawha River (EPA, 1973)................................................ 20
Figure 2. Simplified thermal oxidizer and rupture disc block flow diagram .............................................. 23
Figure 3. 0.5-inch NPS vent/drain pipe and rupture disc ............................................................................ 25
Figure 4. Rupture disc piping and vent pipeline to atmosphere on roof ..................................................... 28
Figure 5. Rupture disc burst sensor post-incident ....................................................................................... 30
Figure 6. New rupture disc.......................................................................................................................... 32
Figure 7. Process data showing sudden pressure decrease when rupture disc burst ................................... 37
Figure 8. Photo of the position of the 1-inch sample line, which had not yet been replaced ...................... 41
Figure 9. The pitting phenomena observed in the small initial hole of the oleum sample line wall ........... 43
Figure 10. The large hole eroded from the outside-in on the oleum sample line........................................ 44
Figure 11. Photo showing the orientation of the small hole to the main hole ............................................. 45
Figure 12. Phosgene shed and full (F) and empty (MT) cylinder locations on day of incident
(not to scale)........................................................................................................................................ 51
Figure 13. SafeAir Phosgene Dosimeter Badge.......................................................................................... 55
Figure 14. The distinct effect of the white plastic adhesive tag on the corrosion of the stainless steel
overbraid ............................................................................................................................................. 63
Figure 15. Damaged hillside phosgene hose removed from phosgene cylinder. The plastic adhesive tag
that covered the damaged section fell off during the hose decontamination procedure. .................... 67
Figure 16. Flex hose comparison photographs: (top to bottom) ruptured riverside hose, flow restricted
hillside hose, a new hose with attached ID tag ................................................................................... 72
Figure 17. Hierarchy of Controls ................................................................................................................ 93
Figure 18. The hose and piping system that supplied phosgene for the release ....................................... 123
Figure 19. ALOHA estimate of phosgene concentrations with MARPLOT GIS overlay ........................ 128

List of Tables
Table 1. Previous rupture disc events in the F3455 unit ............................................................................. 34
Table 2. Phosgene hose change-out frequency ........................................................................................... 65
Table 3. Flexible hoses for phosgene service as listed in the DuPont P3H Standard: Flexible Chemical
Hose for Highly Toxic Services .......................................................................................................... 69
Table 4. Preliminary Risk Assessment by DuPont Engineering, 1988 (Appendix E) ................................ 76
Table 5. PHA enclosure recommendation delays timeline ......................................................................... 78

v
E.I. DuPont de Nemours & Co. Final Report September 2011

Table 6. Sum of Belle plant retirements and new hires from 2005 to 2009................................................ 92
Table 7. DuPont Belle RMP-covered chemicals and threshold quantities ............................................... 102

vi
E.I. DuPont de Nemours & Co. Final Report September 2011

List of Acronyms and Abbreviations


ACC American Chemistry Council
ACGIH American Conference of Governmental Industrial Hygienists
AIHA American Industrial Hygiene Association
ALOHA Area Locations of Hazardous Atmospheres
CERCLA Comprehensive Environmental Response, Compensation, and Liability Act
CFR Code of Federal Regulations
Cl2 chlorine
CMMS Computerized Maintenance Management System
CO carbon monoxide
CSB U.S. Chemical Safety and Hazard Investigation Board
DCS distributed control system
DMA dimethylamine
DMS dimethylsulfate
ECF ethyl chloroformate
EMS emergency medical services
FRC flame-resistant clothing
EPA U.S. Environmental Protection Agency
GIS Graphical Information System
HCl hydrochloric acid
HTM Highly Toxic Materials
IDLH immediately dangerous to life and health
KCEAA Kanawha County Emergency Ambulance Authority
KPEPC Kanawha-Putnam County Emergency Planning Committee
LDAR Leak Detection and Repair
MIC methyl isocyanate
MM million (old notation style)
MOC Management of Change
NDE non-destructive examination
NIMS National Incident Management System
NIOSH National Institute for Occupational Safety and Health
NOAA National Oceanic and Atmospheric Administration
NPS nominal pipe size

vii
E.I. DuPont de Nemours & Co. Final Report September 2011

OSHA U.S. Department of Labor, Occupational Safety and Health Administration


OTPT Oleum Tower Pump Tank
PEL Permissible Exposure Limit
PHA Process Hazard Analysis
PM Preventive Maintenance
ppm parts per million
psig pound-force per square inch gauge
PSSR pre-startup safety review
PSM OSHA Process Safety Management Standard (29 CFR 1910.119)
PTFE polytetrafluoroethylene
RCRA Resource Conservation and Recovery Act
RMP Risk Management Plan
RQ reportable quantity
SAP System Application & Products
SAR Spent Acid Recovery Unit
SCBA self-contained breathing apparatus
SLM Small Lots Manufacturing Unit
SOPs Standard Operating Procedures
TQ threshold quantity
TWA time-weighted average
VOC volatile organic compounds

viii
E.I. DuPont de Nemours & Co. Final Report September 2011

Executive Summary
On January 22 and 23, 2010, three separate incidents at the DuPont plant in Belle, WV, involving releases

of methyl chloride, oleum, and phosgene, triggered notification of outside emergency response agencies.

The incident involving the release of phosgene gas led to the fatal exposure of a worker performing

routine duties in an area where phosgene cylinders were stored and used.

Operators discovered the first incident, the release of methyl chloride, the morning of January 22, 2010,

when an alarm sounded on the plants distributed control system monitor. They confirmed that a release

had occurred and that methyl chloride was venting to the atmosphere. Managers assessing the release

estimated that more than 2,000 pounds of methyl chloride may have been released over the preceding 5

days.

The oleum release, the second incident, occurred the morning of January 23, 2010. Workers discovered a

leak in an overhead oleum sample pipe that was allowing a fuming cloud of oleum to escape to the

atmosphere. The plant fire brigade, after donning the appropriate personal protective equipment, closed a

valve that stopped the leak about an hour after it was discovered. No injuries occurred, but the plant called

the Belle Volunteer Fire Department to assist.

The third incident, a phosgene release, occurred later that same day when a hose used to transfer

phosgene from a 1-ton cylinder to a process catastrophically failed and sprayed a worker in the face while

he was checking the weight of the cylinder. The employee, who was alone when exposed, was assisted by

co-workers who immediately responded to his call for help. Initial assessments by the plants

occupational health nurse indicated that the worker showed no symptoms of exposure prior to transport to

the hospital for observation and treatment. A delayed onset of symptoms, consistent with information in

phosgene exposure literature, occurred after he arrived at the hospital. His condition deteriorated over the

next day and he died from his exposure the next night.

9
E.I. DuPont de Nemours & Co. Final Report September 2011

At the request of the Board, the U.S. Chemical Safety and Hazard Investigation Board (CSB)

investigation team examined all three incidents at Belle due to the severity and potential for even greater

consequences and to understand how and why they could occur at a DuPont facility. DuPont is regarded

as an industry leader in the advancement of health and safety practices and develops sound, respected, and

widely used safe practice guidance. With such a reputation, the CSB was interested in examining the

conditions at the Belle facility that led to a decline in adherence to the higher standard of performance that

the corporation historically held.

The CSB incident investigation determined root and contributing causes for each of the three incidents.

An overall analysis revealed common deficiencies in the following management systems:

Maintenance and inspections


Alarm recognition and management
Incident investigation
Emergency response and communications
Hazard recognition

The CSB found that each incident was preceded by an event or multiple events that triggered internal

incident investigations by DuPont, which investigated all of these precursor events and issued

recommendations and corrective actions. Despite investigating these preceding events, the

recommendations and corrective actions did not prevent the occurrence of similar events.

Because of recent changes to the Kanawha County Metro 9-1-1 response policies and procedures that

could lead to delays in treatment for future incidents, the CSB investigators also examined concerns

raised by the emergency response organizations. These concerns included the timeliness and quality of

10
E.I. DuPont de Nemours & Co. Final Report September 2011

information provided to dispatchers and EMS personnel who responded to two of the incidents and which

mirrors issues identified in the CSB Bayer CropScience August 2008 incident investigation. 1

The CSB identified the following root causes:

Methyl Chloride Incident (January 22, 2010, 5:02 a.m.)


DuPont management, following their Management of Change process, approved a design for the

rupture disc alarm system that lacked sufficient reliability to advise operators of a flammable

methyl chloride release.

Oleum Release Incident (January 23, 2010, 7:40 a.m.)


Corrosion under the insulation caused a small leak in the oleum pipe.

Phosgene Incident (January 23, 2010, 1:45 p.m.)


DuPonts phosgene hazard awareness program was deficient in ensuring that operating personnel

were aware of the hazards associated with trapped liquid phosgene in transfer hoses.

DuPont relied on a maintenance software program that was subject to changes without

authorization or review, did not automatically initiate a change-out of phosgene hoses at the

prescribed interval, and did not provide a back-up process to ensure timely change-out of hoses.

DuPont Belles near-miss reporting process was not rigorous enough to ensure that the near

failure of a similar phosgene transfer hose, just hours prior to the exposure incident, would be

immediately brought to the attention of plant supervisors and managers.

DuPont lacked a dedicated radio/telephone system and emergency notification process to convey

the nature of an emergency at the Belle plant, thereby restricting the ability of personnel to

provide timely and quality information to emergency responders.

1
CSB-2008-I-WV (Bayer CropScience).

11
E.I. DuPont de Nemours & Co. Final Report September 2011

The CSB makes recommendations to


Occupational Safety and Health Administration (OSHA)

DuPont Belle, WV, plant

E.I. DuPont de Nemours & Co., Inc.

Compressed Gas Association of America (CGA)

American Chemistry Council (ACC) Phosgene Panel

12
E.I. DuPont de Nemours & Co. Final Report September 2011

1.0 Introduction

1.1 Background
At 5:02 a.m. on Friday, January 22, 2010, a release of methyl chloride activated an alarm in the F3455

unit control room, signaling the first of three incidents that would occur over the next 33 hours at the

DuPont Belle, WV, facility. No injuries were associated with this incident, but the release went

undetected for as long as 5 days. DuPont estimates that more than 2,000 pounds of methyl chloride

released to the atmosphere.

At 7:40 a.m. on Saturday, January 23, 2010, a contractor reported seeing a fuming plume on a 1-inch

diameter sample pipe in the Spent Acid Recovery (SAR) unit. Operations personnel confirmed that oleum

was leaking; thus, a fume alert was activated for the entire Belle plant. Plant fire brigade members

responded to the release and closed valves that stopped the leak at about 8:09 a.m., after which the all

clear was sounded.

The third incident occurred just 6 hours later. At approximately 1:45 p.m., an operator walked into the

phosgene cylinder storage area in the Small Lots Manufacturing (SLM) unit and was sprayed in the face

and upper torso with phosgene when a flexible hose suddenly ruptured. The worker called for assistance

and coworkers immediately went to his aid. His personal dosimeter indicated that he had been exposed to

a significant dose of phosgene; however, he did not exhibit immediate signs of breathing problems. About

3 hours after arriving at the hospital his condition deteriorated, and he died the following night.

No injuries occurred as a result of the first two releases, but communication to Metro 9-1-1 dispatchers

regarding the nature of each release on Saturday became an issue post-incident. The CSB investigators

examined how information related to the incidents was conveyed to Metro 9-1-1 dispatchers. The CSB

also interviewed Kanawha County Ambulance Authority (KCEAA), Kanawha-Putnam Emergency

Planning Committee (KPEPC), and Metro 9-1-1 representatives to assess each incident and determine if

13
E.I. DuPont de Nemours & Co. Final Report September 2011

actions could be taken to improve communication methods to prevent recurrence of the issues brought to

the attention of county officials. During the Saturday afternoon call for assistance from DuPont, Metro 9-

1-1 dispatchers were not provided with sufficient information regarding the nature of the emergency and

the chemicals involved to adequately inform responding EMS personnel. Many of those interviewed were

familiar with the role of the CSB, having participated in conferences and interviews as part of the CSB

investigation of the August 2008 Bayer CropScience incident.

Due to recurring communication problems associated with emergency responses to chemical plants in the

Kanawha Valley, responding medical units established a practice of waiting before going onto a property

that called for assistance. EMS personnel respond to a staging area as far as a mile away where they

remain until they receive more detailed information about the material involved and whether the victim

has been, or will need to be, decontaminated prior to transport to a hospital. Emergency response

organizations developed this practice as EMS personnel were receiving information that was sometimes

so imprecise that they could not ensure that they or their equipment would not be contaminated by a

hazardous chemical as a result of transporting an exposed victim.

In examining the activities of employees involved in the response, the CSB learned that two other DuPont

employees were also possibly exposed to phosgene. One worker, after he transported the victim part of

the way to the plant medical center in a company truck, noticed that his dosimeter was discolored,

indicating exposure. The second exposure occurred when a worker, unaware of the phosgene release,

went into the area of the phosgene shed and noticed an odor that he had never smelled before. Unsure of

what the odor was, he left the area and joined his co-workers in the control room.

1.2 Investigative Process


Via the media and the National Response Center (NRC), the CSB monitored and tracked information

related to the chemical release incidents at the DuPont Belle, WV, facility throughout the weekend of

January 22 and 23, 2010. On January 25, 2010, the CSB Board deployed an investigation team. Because

14
E.I. DuPont de Nemours & Co. Final Report September 2011

of the number and potential for more severe consequences at the DuPont Belle plant over this 2-day

period, the CSB launched an investigation to determine the root and contributing causes, which it would

use to issue recommendations to help prevent similar occurrences. Although the consequences of the first

two incidents were not as severe as the third, the CSB decided that since the three incidents occurred in

less than 2 days, including one that led to a fatality, all three would be investigated to determine any

common causes.

The investigative team arrived at the Belle Plant on January 26, 2010, and met with Occupational Safety

and Health Administration (OSHA) inspectors; U.S. Environmental Protection Agency (EPA) officials;

and DuPont representatives to explain the CSBs authority and purpose for conducting the investigation.

The CSB investigation team remained onsite for 2 weeks and subsequently visited Belle to conduct

independent investigations of each of the three DuPont Belle, WV, facility incidents. During its

investigations, CSB investigators

interviewed plant personnel, emergency responders, plant supervisors and managers, and

corporate personnel;

coordinated the examination, removal, and storage of physical evidence;

requested and reviewed relevant documentation;

reviewed technical and industry guidance, standards, and regulations;

discussed emergency response issues with the KPEPC, KCEAA, and Metro 9-1-1 dispatch center

officials;

entered into joint testing protocol agreements with DuPont, OSHA, and the EPA;

observed metallurgical testing of the oleum sample line and the phosgene stainless steel overbraid

hose; and

observed analytical testing and analysis of the polytetrafluoroethylene (PTFE) transfer hoses

involved in the phosgene release.

15
E.I. DuPont de Nemours & Co. Final Report September 2011

1.3 E.I. DuPont de Nemours & Co., Inc.


1.3.1 Company History
E.I. DuPont de Nemours and Co, named after its French founder, Eleuthre Irne du Pont, was

established in 1802 as a gunpowder manufacturing company on the Brandywine River in Wilmington,

DE. DuPont grew as a manufacturer of gunpowder and explosives in the United States and in 1902

transitioned into a science-based chemical company. DuPont established Experimental Station, the first

industrial laboratory where researchers and scientists began work on nitrocellulose chemistry and

smokeless powders to improve military rifles for the World War I effort. By the 1920s, DuPont purchased

several chemical companies and focused on polymers, which led to the discovery of neoprene (synthetic

rubbers), polyester, and nylon by 1935. Many of these products were in demand during the Second World

War. Further work with plastics and fibers led to the development of Teflon, Lucite, Nomex, and

Mylar in the 1950s. DuPont also introduced a number of inorganic insecticides and fungicides such as

Lannate (methomyl) and Telvar, which eventually led to the establishment of its agricultural products

business. By the mid-1980s, DuPont had grown to almost 100 major businesses selling a wide range of

materials such as textiles, agricultural chemicals, petroleum, and biomedical products.

1.3.2 DuPont Business Areas and Corporate Management


DuPont, headquartered in Wilmington, DE, has 58,000 employees in more than 80 countries. The

company offers a broad range of products for industry and consumer use, including pesticides,

electronics, apparel, and biomedical supplies. Five business platforms comprise the DuPont organization:

Agriculture and Nutrition, Coatings and Color Technologies, Performance Materials, Electronics and

16
E.I. DuPont de Nemours & Co. Final Report September 2011

Communications, and Safety and Protection. Within each business platform are strategic business areas 2

focusing on the production, sale, and distribution of products and services related to each marketing area.

The Crop Protection business area, a segment of the Agriculture and Nutrition platform, is responsible for

the development, manufacture, and sale of fungicides, herbicides, insecticides, and seed treatments

globally. The agriculture industry uses DuPont Crop Protection products on a variety of crops worldwide

including cotton, soybeans, fruits, and vegetables. The F3455 and SLM units at the Belle Plant

manufacture intermediate chemicals for their Crop Protection products. In 2009, the Agriculture and

Nutrition platform had the most sales of any business area at $8.3 billion.

A 13-member Board of Directors, including the chairperson and CEO, manage DuPont. Executive

committees made up of board members and representatives from DuPont businesses oversee areas such as

environmental policy, corporate governance, strategic direction, and auditing. In 2010, DuPont had global

sales of $31.5 billion and ranked as the third-largest chemical company in profits and second in revenues

in the world.

1.3.3 Safety at DuPont


Concern for safety and health at DuPont became a part of the companys structure in 1805 due to the

hazards of producing gunpowder and explosives. The early corporate safety program was rooted in

process safety concepts more than a century before governing safety regulations existed. Practices such as

safe siting of buildings, explosion venting concepts, incident investigation processes, and emergency

response were implemented in the DuPont gunpowder mills throughout the 19th century.

2
Pioneer Hy-bred, Crop Protection, Nutrition and Health, Electronics and Communications, Performance Coatings,
Performance Polymers, Protection Technologies, Building Innovations, Sustainable Solutions, Chemicals and
Fluoroproducts, Titanium Technologies, and Applied Biosciences.

17
E.I. DuPont de Nemours & Co. Final Report September 2011

The company continued to focus on health and safety to improve safety performance and in 1915 created

its first corporate safety division, which was responsible for technical training, safety inspections, project

design reviews, and the purchase of safety equipment. According to DuPont incident records, the safety

division participation in facility operations decreased incident rates throughout the company. As a result,

individual sites established site-specific safety groups in the mid-1930s. Hazard elimination was

recognized as a priority above education and personal protection (Klein, 2009).

1.3.3.1 Early Process Safety Program


The release of highly toxic methyl isocyanate (MIC) at the Union Carbide Corp. in Bhopal, India, resulted

in nearly 3,800 immediate deaths, and 16,000 are estimated to have since died as a result of exposure,

while more than 100,000 still report associated illnesses. In response to the Union Carbide incident,

chemical companies, industry associations, and government agencies directed efforts to decrease process

safety risks, which eventually led to the establishment of the OSHA Process Safety Management (PSM)

Standard (29 CFR 1910.119), EPA Chemical Accident Prevention Program, and the creation of the CSB

as part of the Clean Air Act amendment of 1990.

Prior to establishing the OSHA PSM Standard, DuPont was practicing many process safety concepts at its

facilities as part of the DuPont Process Hazards Management (PHM) Program. After a 1965 incident in

Louisville, KY, killed 12, the company directed all sites to perform hazard reviews to evaluate the safety

of site processes, which eventually became a corporate Process Hazards Review (PHR) program. The

PHR was intended to prevent serious process-related incidents, and each site handling hazardous

substances had to have a PHM program.

The Bhopal incident contributed to an increase in DuPonts focus on PHM, particularly in the

manufacture of MIC. DuPont developed an inherently safer method of manufacturing and handling MIC

that eliminated MIC bulk storage, as it relied on producing and directly consuming MIC. The company

also created the Highly Toxic Materials (HTM) Subcommittee to review the global management of toxic

18
E.I. DuPont de Nemours & Co. Final Report September 2011

chemicals. In 1985, HTM became a corporate guideline, and a separate subcommittee was established to

focus on each of the 15 highly hazardous materials identified within the company. DuPont continued to

refine its PHM program, eventually developing professional guidance for process safety and OSHA PSM

rulemaking (Mottle et al., 1995).

1.3.3.2 Zero Incidents Goal


DuPont introduced the zero incidents goal in the early 1900s as a management directive to drive injury

rates down to zero through continuous improvement of safety practices. The zero concept became a

core strategy as the company grew and embraced the philosophy that all injuries, occupational illnesses,

and environmental incidents are preventable and that the goal for all is zero.

DuPont became recognized throughout industry as a safety innovator and leader. The company offers

services as a safety resource for other corporations to evaluate and improve workplace safety, which

include methodologies and technical training to manage and improve employee and contractor health and

safety performance as well as process safety improvements.

19
E.I. DuPont de Nemours & Co. Final Report September 2011

1.4 DuPont Belle Plant

Figure 1. DuPont Belle, WV, facility on the Kanawha River (EPA, 1973)

The DuPont Belle plant is located in Belle, WV, about 8 miles east of Charleston, the state capital . The

plant occupies about 723 acres along the Kanawha River and sits in an industrial, commercial, and

residential use area. The plant was established in the West Virginia coal country as part of a post-World

War I effort to produce ammonia. In the early 1920s DuPont spent $27 million 3 on a highly complex

production facility with atmospheric compressors capable of producing 25 tons of ammonia per day.

Belles high-pressure ammonia technology yielded a host of collateral benefits. Methanol was initially

manufactured on a small scale and then rapidly expanded to 1 million gallons a year. By 1935, Belle had

3
Equivalent to $332 million in 2010, according to the Bureau of Labor Statistics Inflation Calculator.

20
E.I. DuPont de Nemours & Co. Final Report September 2011

become DuPont's largest facility with more than 80 different chemical products, which included the first

synthetic urea used in fertilizers and plastics. In 1939, DuPont began producing nylon chemical

intermediates at Belle, and by 1944 the plant was producing 30 million pounds of synthetic polymers per

year. Expansion of nitrogen and nylon intermediate production at Belle continued after the war, and

product lines were introduced regularly. In 1969 Belle began producing the fungicide Benlate. Currently,

the DuPont Belle plant produces a variety of organic chemicals and agricultural intermediates and

products. According to company documents, the plant had the best safety record of any DuPont

production facility prior to the incidents of January 22 and 23. 4

In January 2010, the DuPont Belle plant employed approximately 440 and had seven primary operating

divisions occupying a 105-acre manufacturing area nearly 1 mile long. The DuPont-operated SAR unit

was owned by Lucite International and operated by DuPont employees. The Belle facility is also the site

of the newly constructed Kureha unit, owned by the Kureha Corp. of Japan, which is operated by Kureha

employees on DuPonts Belle site. The Kureha production unit uses glycolic acid produced by DuPont as

a feedstock for polyglycolic acid, a specialty plastic.

The DuPont Belle plant holds a Resource Conservation and Recovery Act (RCRA) Part B Treatment and

Storage Permit for onsite handling of waste materials, in addition to a RCRA-permitted drum storage

facility onsite. The Belle plant participates in a Community Action Council (CAC), comprised of citizens

from neighboring communities and representatives from the industrial facilities in the region, 5 that aims

to address citizen concerns regarding site safety, health, and environmental performance.

4
www.2.dupont.com/heritage.
5
DuPont Belle Plant Information Sheet.

21
E.I. DuPont de Nemours & Co. Final Report September 2011

2.0 Methyl Chloride Release (January 22, 2010)

2.1 Background
The Belle plants F3455 unit manufactures the intermediate F3455, a chemical that is shipped to another

DuPont facility to make the herbicide Velpar. Due to the exothermic reaction in the first reactor,

dissolved methyl chloride vaporizes and normally exits through the reactor vent line along with carbon

dioxide, nitrogen, and trace amounts of dimethylamine (DMA) vapor through a process scrubber and then

to a thermal oxidizer for emission control. To avoid damage to the scrubber 6 if excessive pressures occur,

a piping connection upstream of the vent line is routed to a rupture disc that will burst and allow venting

outside on the roof of the building which contains two reactors (Figure 2). However, due to a lack of

safety considerations during installation, a 0.5-inch weep hole 7 was placed on the vent line inside the

building; consequently, dangerous chemicals vent inside the building if the rupture disk bursts.

Unaware that the rupture disc had blown during a nitrogen purge activity before the reactor startup, plant

personnel proceeded with the normal production run. For nearly 5 days, methyl chloride vapor passed

through the blown rupture disc and escaped into the operation building and outside atmosphere. On the

fifth day, the methyl chloride vapors interfered with the chemical sensor configured to detect ethyl

chloroformate (ECF), which alerted the workers.

6
A thermal oxidizer is a process unit for air pollution control in many chemical plants that decomposes hazardous
gases at a high temperature and releases them into the atmosphere.
7
In process vent lines that lead to the atmosphere, protection must be installed to prevent ambient moisture -- from
rain or other elements -- from collecting within the vent line. One such protection is a weep hole, a small hole
drilled into a vent line that allows drainage.

22
E.I. DuPont de Nemours & Co. Final Report September 2011

Figure 2. Simplified thermal oxidizer and rupture disc block flow diagram

2.1.1 Methyl Chloride


Methyl chloride, also called chloromethane or monochloromethane, is a colorless gas with a faint sweet

odor at low concentrations. 8 The odor may not be noticeable and cannot be relied upon as warning of

concentrations that are dangerous to health. 9 Methyl chloride is extremely flammable; has a potent

8
The odor threshold, or concentration, of methyl chloride detectible by most humans varies between 10 and 250
ppm.
9
http://www.oxy.com/Our_Businesses/chemicals/Documents/methyl_chloride/Methyl%20Chloride%20Handbook
.pdf.(11/2009)

23
E.I. DuPont de Nemours & Co. Final Report September 2011

narcotic effect similar to trichloromethane, also known as chloroform; and is listed as a Group 3

carcinogen 10 by the International Agency for Research on Cancer (IARC). The OSHA 8-hour time-

weighted average (TWA) concentration is 100 ppm and the National Institute of Occupational Safety and

Health (NIOSH)-designated Immediately Dangerous to Life and Health (IDLH) concentration is 2,000

ppm.

Symptoms of methyl chloride exposure include dizziness, confusion, and nausea, and at higher

concentrations, extreme nervousness, trembling, and possible loss of consciousness. High concentrations

or long exposure can be fatal. The gas is also heavier than air and therefore settles close to the ground.

2.2 Incident Description


The F3455 process was in the first series of batch runs following an extended maintenance outage from

September 12, 2009, through January 17, 2010. The release is thought to have initiated on January 17

during the first batch run in the unit and continued until discovered on January 22; the release rate may

have been sporadic throughout this period.

On January 22, 2010, an air monitor alarm on the process control monitor alerted plant operating

personnel of a chemical release while they were adding DMA 11 to the reactor. The sensor for this alarm,

located on the third floor of the F3455 building, is calibrated to activate when it detects ECF at 0.5 ppm.

The methyl chloride vapors interfered with the ECF sensors on the third floor and activated the alarm.

The distributed control system (DCS) recorded the alarm at 5:02 a.m., and responding operators saw a

diffused fog and a liquid puddle near a 0.5-inch nominal pipe size (NPS) vent/drain pipe referred to as a

10
Substances the IARC lists as Group 3 carcinogens are mixtures or agents for which evidence of carcinogenicity in
humans is inadequate and limited in experimental animals.
11
DMA is a toxic and extremely flammable, colorless product with a fishy or ammonia-like odor. DMA attacks the
respiratory system and irritates eyes and skin and at higher concentrations can cause pulmonary edema. The
OSHA 8-hour TWA is 10 ppm and the NIOSH IDLH is 500 ppm. Humans can detect DMA odors at 0.34 ppm
(Sittig, 2008). DMA is a heavier than air vapor and settles close the ground

24
E.I. DuPont de Nemours & Co. Final Report September 2011

weep hole (Figure 3). This connection was associated with a thermal oxidizer vent stack, that vents to

the atmosphere on the roof of the building during a process upset. Operators notified the board operator at

5:19 a.m. when they found the source of the release.

Figure 3. 0.5-inch NPS vent/drain pipe and rupture disc

2.2.1 ECF Sensor Alarm


The ECF sensor was detecting chlorine, not ECF. The ECF sensor is responsive to chemicals composed

of chlorine (i.e. ethyl-chloroformate [ECF] and methyl-chloride); consequently, on the fifth day, the

chlorides in the release were of sufficient concentration near the ECF sensor to activate the alarm.

2.2.2 Odor Detection Considerations


The methyl chloride, DMA, and hydrochloric acid (HCl) mixture is extremely odorous; however, due to

the nature of the F3455 process, operating personnel would have had to be in the area of the 0.5-inch

weep hole at the time of the release to see or smell the leak.

Methyl chloride liberated during this phase of the reaction would have likely taken the normal route to the

thermal oxidizer piping, where it would have been consumed and vented to the atmosphere unnoticed.

25
E.I. DuPont de Nemours & Co. Final Report September 2011

The vent releases products of the reaction into the room if a rupture disc is blown and if the pressure

inside the pipe is greater than the pressure in the room.

The rupture disc piping was routed to the atmosphere above the roof of the building, which would have

provided an outlet path for the methyl chloride vapor where it would have dissipated and dispersed

without notice.

The day before the leak was discovered, a crew performed a leak detection and repair (LDAR) 12

inspection on the third floor of the building near the location of the release. The volatile organic

compounds (VOC) electronic monitor was calibrated to detect methyl chloride, ECF, DMA, and

methanol. Although an area within 12 inches of the weep hole was checked for leaks with the monitor, it

did not detect any VOCs.

2.2.3 Incident Response


In response to the ECF alarm, operators using a VOC analyzer to search for the source of the vapor

immediately smelled an offensive odor on the third floor. They saw steam-like fumes near the vent pipe

and dripping liquid puddling on the floor, both clear indications that the rupture disc had burst (Figure 4).

They left the process area, closed all valves leading to the vent line, and cooled the reactors to stop the

process. At about 9:30 a.m., maintenance mechanics replaced the rupture disc and burst sensor.

After receiving confirmation of the release, the board operator notified the process supervisor who then

calculated the estimated duration and magnitude of the release. After performing these calculations, the

supervisor notified the plant manager, the Safety Health and Environmental (SHE) manager, the area

manager, and the unit technology leader and told them that the release may have been ongoing for the

12
The Clean Air Act requires refineries and chemical plants to develop and implement an LDAR program to control
fugitive emissions, which occur from leaks in valves, pumps, compressors, pressure relief valves, flanges,
connectors, and other piping components.

26
E.I. DuPont de Nemours & Co. Final Report September 2011

entire run of nine batches, which occurred over 5 days. DuPont estimated that approximately 2,000 13

pounds of methyl chloride were likely released to the atmosphere.

During the initial phases of the DuPont incident investigation, employees discovered that the burst sensor

on the rupture disc had started alarming 5 days prior to the incident. Due to its history of unreliability,

operators likely became desensitized to this alarm. The burst sensor was the first in this sequence of

incidents that led to a safety pause 14 at the plant.

13
DuPont, in its final investigation report, determined that 2,045 pounds of methyl chloride and 25 pounds of HCl
released to the atmosphere as a result of this incident.
14
A safety pause is a structured work stoppage that the plant manager initiates to engage the entire workforce with
the objectives of increasing awareness of hazards, providing safety education, and addressing past incidents. A
safety pause was initiated at the Belle facility on Saturday, January 23, 2010, because of the incidents at the
F3455 and SAR units.

27
E.I. DuPont de Nemours & Co. Final Report September 2011

Figure 4. Rupture disc piping and vent pipeline to atmosphere on roof

28
E.I. DuPont de Nemours & Co. Final Report September 2011

Once DuPont determined that the release quantity exceeded the Comprehensive Environmental Response,

Compensation and Liability Act (CERCLA) reportable quantity (RQ) of 100 pounds, 15 in compliance

with CERCLA of 1980 it reported the release of methyl chloride to the NRC and to the West Virginia

State Department of Homeland Security Emergency Operations Center, which notified the U.S. Coast

Guard. Kanawha County Metro 9-1-1 was not informed of the release until 2:00 p.m. on January 22,

2010, 9 hours after discovery.

2.2.4 Community Impact


DuPont estimated that between January 17 and 22, 2010, 2,045 pounds of methyl chloride; 25 pounds of

hydrogen chloride; and trace amounts of DMA released to the atmosphere through a vent line on the roof

of the F3455 building. No monitoring information was available to determine the concentrations of

chemicals released to the atmosphere through the vent line. If monitoring information had been

recorded, a more accurate estimate of chemical concentration would have provided data about when the

release started and the potential for offsite impact. No workers at the facility reported symptoms from

methyl chloride or any of the other toxic chemicals either during or after the release. DuPont did not

receive any odor complaints from the community.

2.3 Incident Analysis


2.3.1 Mechanical Integrity
Rupture discs are overpressure protection devices used in processes operating above ambient pressure and

are intended to prevent equipment damage, including catastrophic failure. Without them, a process upset

can cause unsafe pressure levels and an overpressure incident. Since these devices activate only when a

system has had an overpressure event, it is imperative that their activation be discovered. In this

15
Under CERCLA, operators of facilities and vessels are required to immediately report releases to the NRC above
the EPA RQ.

29
E.I. DuPont de Nemours & Co. Final Report September 2011

application, the rupture disc releases hazardous chemicals to the atmosphere. One approach to help with

early detection is to evaluate the alarm management process and, where appropriate, adjust process

parameters so that an alarm will activate prior to the disc actually bursting. Another is to evaluate the

process and eliminate the conditions that increase the pressure that cause the disc to burst. Regardless,

once systems have been selected, the configuration should be reviewed by a team, including process

engineers, control engineers, and operations managers (Lees, 2005).

Figure 5. Rupture disc burst sensor post-incident

DuPont Belle used a burst sensor intended to notify the board operator that the rupture disk (Figure 5)

activated. A burst sensor is a thin plastic membrane with embedded wires installed on top of the rupture

disc. Small electrical current passes through the wires. When the rupture disc activates, the membrane and

embedded wires break, triggering the alarm.

The CSB learned that the rupture discs and sensors associated with this system were historically

problematic. The burst sensor involved in the January 22, 2010, incident had been replaced many times

30
E.I. DuPont de Nemours & Co. Final Report September 2011

because it was unreliable. Initially the sensor was battery-operated, sending signals to a remote receiver in

the control room rather than to the process control monitor. However, the battery life was short;

consequently, operators received frequent false, or nuisance, alarms. According to Management of

Change (MOC) documentation, the burst sensor [was] in and out of alarm every 3 minutes and required

replacements almost monthly. When its batteries failed, the transmitter sent an alarm to the remote

receiver to notify the operators. The receiver displayed the same alarm text as when the sensor detected a

burst rupture disc. Because the batteries needed frequent replacing and because the operators had to wait

for an electrician to change the batteries, the false alarms became a nuisance.

Battery life, however, was not the only reported shortcoming of burst sensors. Operators told the CSB

investigators that burst sensors were so delicate that they could sometimes tear during installation and that

liquid condensation on top of the sensors sometimes caused them to fail and trigger a false alarm.

An improved burst sensor was installed on the DCS while the unit was down for maintenance just before

the incident. Operators indicated they were not retrained to respond to the more reliable burst sensor

alarm and still considered it a nuisance.

2.3.2 Design and Maintenance of Rupture Discs


The rupture disc involved in the incident was a 4-inch diameter graphite rupture disc, designed to rupture

at 15 psig, and mounted in neoprene casing (Figure 6). While the rupture disc is on a preventive

maintenance (PM) schedule, the annual inspection was so infrequent that the disc is replaced only when it

has activated or is removed for certain processes. Operators told the CSB investigators that once removed,

the rupture discs, intact or compromised, are discarded and replaced with new ones. Even without a burst

31
E.I. DuPont de Nemours & Co. Final Report September 2011

sensor, all overpressure protection devices, including rupture discs, should be routinely checked on an

effective PM schedule as a layer of protection. 16

Figure 6. New rupture disc

2.3.3 Previous Incidents of Rupture Discs Bursting


From 2005 to 2010, the rupture disc on the F3455 unit vent line experienced nine recorded activations

(Table 1). On April 11, 2006, the rupture disc activated three times. DuPont determined that the disc was

most likely experiencing thermal or hydraulic shock. Thermal shock would occur from boiling reactor

16
BS&B Safety Systems, Inc, Special Applications and Preventive Maintenance, Catalog 77-1007, Section B.

32
E.I. DuPont de Nemours & Co. Final Report September 2011

vapor mixing with cool liquid on the disc due to its close proximity to the reactor; hydraulic shock would

occur from any sloshing in the line upstream of the disc. This recurring problem was remedied by

moving the rupture disc farther away from these units and eventually to the third floor toward the extreme

end of the vent line.

On May 6, 2006, a rupture disc activation at the same facility went unnoticed for 48 hours, which

illustrates how the January 22, 2010, release could have gone undetected for 5 days. In the May incident,

although operators complained about strong odors in the F3455 building, the rupture disc was never

considered as the source; indeed, operators and supervisory staff identified multiple locations where

fugitive emissions could have produced the offensive smell. Eventually, when a new batch of F3455 was

started, an operator near the vent line saw the rupture disc fuming, indicating that it was the odor source.

At the Belle facility, pipe blockage at the unit was the most commonly reported cause of premature

rupture disc activation (Table 1). The F3455 process creates various solids in the vent and process lines,

which eventually block flow, increasing the pressure in the system. Once the blockage is melted by the

process temperature or forced through the line due to the increased pressure, the resulting pressure spike

activates the rupture disc.

33
E.I. DuPont de Nemours & Co. Final Report September 2011

Previous Rupture Disc Incidents


Date Cause
05/20/05 Unknown
05/31/05 Pressure Control Issues
04/11/06 Hydraulic/Thermal Shock 17
05/06/06 Blockage 18
06/16/06 Unknown
05/30/07 Ruptured during Water Cleaning
06/12/07 Blockage
04/15/08 Blockage
02/24/09 Blockage

Table 1. Previous rupture disc events in the F3455 unit

2.3.4 Management of Change--Technology and Subtle Change


Within DuPont, MOC procedures are defined at a corporate level and adopted according to each sites

procedures. At the corporate level, the PSM Standard defines two types of MOC: technology (MOC-T)

and subtle changes. MOC-T is defined as a change in hazards of materials (including the introduction of

chemicals), a change in equipment design basis, or a change to the process design basis. Subtle changes

are defined as any change within the documented [process technology] that is not a replacement in

kind. 19 Regarding high-hazard processes, such as the F3455 and SLM units at Belle, the corporate PSM

Standard states, [S]ubtle changes in the field can (and have) led to catastrophic events. However, even

with this knowledge the MOC team at Belle incorrectly categorized the burst sensor installation as a

subtle change.

17
This incident was actually three incidents over a short period. The rupture disc was discovered ruptured and
replaced three times before the unit was shut down for further investigation.
18
This incident went undiscovered for 48 hours.
19
The corporate DuPont PSM Standard defines replacement in kind as the replacement of an instrument or
electrical, piping, or other process equipment component with an identical part or an approved equivalent part that
is specified by the applicable DuPont Engineering standard.

34
E.I. DuPont de Nemours & Co. Final Report September 2011

At the Belle site the standard operating procedures (SOPs) do not distinguish between MOC subtle

changes and MOC-T. The MOC package documentation, however, shows that subtle, often referred to as

minor, changes are not subjected to the same in-depth review as a MOC-T. When the MOC is marked

as subtle, the level of safety review is at the discretion of the MOC team leader.

The MOC package that first installed the rupture disc burst sensor was marked as a subtle change and

included a What If review that stated, What if you get a false positive indication (indicating failed

disc, but not actually failed)? Not a safety issue. Shut down and investigate.

This type of review did not go deep enough to confirm that false-positives could lead to nuisance alarms,

which can create risk by desensitizing operators to a hazard and be more detrimental than the absence of

the alarm. In the MOC section marked Reason for this Type of Safety Review, the response by the

MOC team leader was Minor Change.

The MOC package that converted the burst sensor from battery-powered to a supplied power device was

also marked as a subtle change. Again, the MOC team leader recorded in the documentation that a What

If review [was] appropriate for the afore-mentioned [sic] change. The MOC did not address the

operators non-battery related concerns for the burst sensor or how to re-train the board operator to no

longer treat the burst sensor alarm as a false-positive.

Because MOC packages deemed subtle are not given the same level of review as MOC-T packages, the

subtle change MOC packages did not identify or prevent the potential causes of this incident.

2.3.5 F3455 Unit Turnaround


On June 6, 2009, nearly 2 years after installing the battery-operated transmitter, DuPont attempted to

eliminate the false alarms caused by low batteries by wiring the transmitter to a standard electrical circuit.

During a unit shutdown that lasted from September 12, 2009, through January 17, 2010, there was

significant maintenance activity, including work that, by its nature triggered alarms; however, these

35
E.I. DuPont de Nemours & Co. Final Report September 2011

alarms did not require response from the operators because there were no live process streams that

would initiate an actual alarm.

DCS data recorded during the shutdown indicated that the pressure in the reactor system increased slowly

from December 18, 2009, to December 20, 2009, when it exceeded the rupture disc rating (Figure 7). The

source of the pressure was a nitrogen valve on a level indicator that slowly leaked nitrogen into the

system. 20 The rupture disc burst, triggering an alarm, as it should have. Under normal, live operating

conditions, the operators would have investigated to understand, acknowledge, and correct the alarm

condition. However, extensive maintenance work was still underway in the unit; thus, the operators did

not address the alarm as they would have under normal operation.

20
The level instrument measures the difference between the pressure in the vapor space inside the top of the reactor
and the pressure under the liquid at the bottom of the reactor. Based on the pressure difference, the control
computer calculates the amount of liquid in the reactor. The nitrogen provides a chemical barrier between the
reactor liquid and the level instrument.

36
E.I. DuPont de Nemours & Co. Final Report September 2011

18

16

14

12

10

8
Pressure

0
12/18/09 12:00 12/18/09 12:00 12/19/09 12:00 12/19/09 12:00 12/20/09 12:00 12/20/09 12:00 12/21/09 12:00 12/21/09 12:00
AM PM AM PM AM PM AM PM

Time

Figure 7. Process data showing sudden pressure decrease when rupture disc burst

The operators did not address the alarm when it triggered in December because they knew that work in

the area was causing nuisance alarms; however, when the ECF alarm activated on January 22, 2010,

operators responded. The board operator in the F3455 control room investigated and observed that the

original alarm from December 21, 2009, was still displayed; the first item on the alarm screen had not

been acknowledged because they had become accustomed to nuisance alarm conditions. 21

21
Under normal operating conditions, when an alarm point activates it will remain in an activated state until the
alarm condition is cleared and acknowledged.

37
E.I. DuPont de Nemours & Co. Final Report September 2011

2.3.6 Second-Party Process Safety Management Audit


In 2007, an audit team of engineers and safety and health experts from other DuPont facilities conducted a

4-day second-party audit 22 of the Crop Protection business at Belle, which included the F3455 and SLM

units. The four-member team audited the units against PSM focus areas such as MOC-subtle change, pre-

startup safety reviews (PSSRs), training, PHA, mechanical integrity, and process technology. While

auditing the F3455 unit and during a review of site and area management practices, the team noted the

many active alarms in the unit control room: [The] control system is not engineered to eliminate alarms

from idled and secure process equipment [and as a result] the contribution to nuisance alarms is

unknown. The audit team recommended that Belle evaluate the control system and develop an

engineered solution to reduce the number of active alarms and establish a policy reflective of

improvements to safely manage operations with active alarms.

During another review of SOPs and worksite practices, the team noted that the Crop Protection procedure

for operating with active alarms did not effectively address alarm activations from idle equipment: The

current situation can lead to human factors errors such as failing to recognize an alarm and misidentifying

an alarm. The team recommended that Belle conduct an engineering evaluation to determine changes

that could separate alarms on active processes from those associated with shutdown equipment so that

operators could readily identify abnormal process conditions.

Both recommendations, added to a corrective action tracking plan, were completed in fourth quarter 2008,

months beyond the original target completion dates. Despite these recommendations, F3455 unit

personnel continued to restart the unit while the alarm was activated, failing to recognize the impact of the

burst sensor alarm.

22
A second-party audit is an independent assessment of PSM systems performed against the requirements of the
DuPont corporate PSM standard.

38
E.I. DuPont de Nemours & Co. Final Report September 2011

2.4 Key Findings


1. The rupture disc alarm system being monitored by a battery-powered transmitter, with batteries

requiring almost monthly replacement, was designated as PSM-critical 23 equipment by DuPont.

2. DuPont ran the equipment with an unreliable battery-powered transmitter for 18 months before

executing a MOC package to convert to a wired power supply.

3. Operators expected maintenance work to trigger alarms, but planning and communication were

insufficient to distinguish which alarms needed immediate attention during the turnaround and

after work was completed.

4. Despite repeated incidents of rupture discs bursting, DuPont did not adequately address the cause

to prevent recurrence.

5. The alarm from the transmitter did not distinguish between a condition that required immediate

attention (ruptured disc burst) and a lower priority condition such as failed batteries.

6. Operators became desensitized to the rupture disc burst alarm.

2.5 Root Causes


1. DuPonts MOC process approved a design for the rupture disc alarm system that lacked sufficient

reliability for minimizing the release of methyl chloride.

2. DuPont did not resolve the nuisance alarm condition in a timely manner despite various safety

reviews.

23
PSM-critical is defined in DuPont SHE Standards S21A and S24 A as components, equipment, or systems whose
failure could cause, allow, or contribute to process incidents that result in death or serious injuries, significant
property damage, or significant environmental impact.

39
E.I. DuPont de Nemours & Co. Final Report September 2011

3.0 Oleum Release (January 23, 2010)

3.1 Background
Lucite International owned the sulfuric acid recovery (SAR) unit on DuPonts Belle plant property and

DuPont employees operated the equipment. The SAR unit produced oleum, which is a solution of sulfur

trioxide dissolved in sulfuric acid. As the sulfuric acid is consumed, the sulfur trioxide converts to

sulfuric acid.

The process unit adjacent to the SAR unit used the oleum to produce methacrylic acid, an ingredient for

acrylic polymers, and then returned the spent oleum to the SAR unit. The SAR unit burned off the

impurities from the spent oleum and used the remaining sulfur compounds to produce clean oleum.

As a result of an unrelated, earlier inspection, the EPA ordered the Belle facility to upgrade emissions

monitoring equipment or improve abatement capacity in the SAR unit. As part of a consent decree with

the EPA issued on April 24, 2009, Lucite International chose to permanently shut down the plant. The

complete and final shutdown of the SAR was concluded in March 2010.

3.2 Incident Description


On January 23, 2010, at about 7:40 a.m., contract personnel working near the SAR unit saw an unusual

cloud near the oleum tower and reported a fume release to the board operator. The contractors estimated

the release to be about midway along the length of a1-inch diameter insulated pipe between the Oleum

Tower Pump Tank (OTPT) and a sample station (Figure 8). The board operator asked the plant operator

to go to the area of the reported leak to determine the nature of the release. The plant operator confirmed

that a leak had developed on the sample piping between the OTPT and the sample station and alerted

other workers in the vicinity to move to a safe area. Based on the information the plant operator provided,

40
E.I. DuPont de Nemours & Co. Final Report September 2011

at about 7:45 a.m. the board operator notified the main gate guard, who then activated a fume alert 24 to

notify the facility of the release.

Figure 8. Photo of the position of the 1-inch sample line, which had not yet been replaced

24
Each plant in the facility has a pre-determined unique number of rings that identify it in case of a release or
emergency.

41
E.I. DuPont de Nemours & Co. Final Report September 2011

A cloud of steam and sulfuric acid mist from this release is reported to have traveled west and dissipated

in an adjacent operating unit. A concrete dike surrounding the OTPT contained liquid from the leak.

There were no reports of exposure to any DuPont or contract employees or the public.

3.2.1 Incident Response


When the plant activates a fume alert, a klaxon bell notifies plant personnel of the location of the incident.

This action also initiates a response by plant fire brigade personnel who go to the facilitys fire station to

obtain the plant fire engine and personal protective equipment (PPE) necessary to respond to the incident.

At about the same time the fume alert was sounded, the gate guard called Metro 9-1-1. The shift

supervisor radioed the gate guard to notify the Belle Volunteer Fire Department, which then dispatched

three engines to the plant. Two of the engines staged outside the plants gate while the third went into the

plant to stand by.

DuPont fire brigade members arrived at the site of the release and set up a water fog spray from the

DuPont fire engine and an oscillating water spray from a nearby hydrant for about an hour. After donning

an acid suit and self-contained breathing apparatus (SCBA), one responder entered the area and closed a

valve, which stopped the release at about 8:09 a.m. The gate guard sounded the all clear at about 8:27

a.m. Calculations estimate that 22 pounds of 20 percent oleum was released during the incident. 25

3.3 Incident Analysis


3.3.1 Reconstructive Analysis
The CSB investigators documented the analysis of the oleum sample line, which was conducted by an

independent metallurgical lab, to determine the incident cause.

25
20 percent oleum has an acid content that is 20 percent greater than pure sulfuric acid.

42
E.I. DuPont de Nemours & Co. Final Report September 2011

Caused by an unknown defect, oleum corroded through a small section of the pipe involved in the release

on January 23, 2010. Starting as a pitting phenomena and finishing slightly larger than a pin hole, the

corrosion penetrated the insulated stainless steel sample pipe (Figure 9).

Figure 9. The pitting phenomena in the small initial hole of the oleum sample line wall

Once oleum was present on the exterior of the oleum pipe, it readily corroded the insulation and steam

tracing line and then created a leak in the steam tracing, causing the steam and oleum to mix. This

reaction created a strong solution of sulfuric acid that rapidly and effectively corroded the stainless steel

sample line exterior, until a second larger hole developed at a location near the original small leak. The

second hole clearly shows corrosion occurring from the outside-in (Figure 10).

43
E.I. DuPont de Nemours & Co. Final Report September 2011

Figure 10. The larger hole eroded from the outside-in on the oleum sample line 26

When DuPont removed the oleum-soaked insulation and cover, a larger hole was visible; the acid had

also corroded a large amount of the steam tracing. When the sample line was properly cleaned, inspection

revealed that the smaller hole was only a few inches away from the larger hole, and after thorough

examination, metallurgists concluded that the small hole in the sample line initiated the oleum release

(Figure 11).

26
Because the oleum pipe was held as evidence, its decontamination was delayed; the size of the holes may have
marginally increased from continued corrosion prior to examination.

44
E.I. DuPont de Nemours & Co. Final Report September 2011

Figure 11. Photo showing the orientation of the small hole to the main hole

3.3.2 Pipe Testing and Analysis


The oleum sample line was tested using gamma ray radiography, ultrasonic thickness (UT), and

metallographic analysis. The metallographic analysis confirmed that the sample line was fabricated from

304L stainless steel, one of the few metals approved by the DuPont Piping Standard for this oleum

service.

The radiographic and UT testing showed that the pipe wall had suffered general thinning from corrosion,

which is expected in most piping applications involving corrosive materials. The thinning rate can predict

the service life of the pipe, and in the case of pipes routing corrosive materials, the expectation is that

45
E.I. DuPont de Nemours & Co. Final Report September 2011

roughly 1 to 2 mils 27 will corrode per year. UT testing and radiography revealed that general wall

thinning of the sample line was much less than the predicted 1 to 2 mils per year and showed much less

thinning than expected for its lifetime. This sample line had been in place for 19 years, which is not

unusual for this type of service.

Only one anomaly, later deemed the initiator of this incident, was found during the testing. During visual

inspection a small hole was discovered 90o off and a few inches from the larger hole. Under microscopic

examination, the small hole corrosion phenomenon could clearly be seen; however, its exact cause is

unknown. One theory is that this small hole may have originated from some sort of manufacturing defect,

but the size and shape of the pitting phenomenon suggest that if this were a manufacturing defect, the

pitting would have occurred around the circumference of the pipe or along the longitudinal axis. This

particular phenomenon does not fall into any easily defined defects. Due to the small size of this pitting, it

is unlikely that routine non-destruction examination (NDE) techniques would have identified this defect.

3.3.3 Previous Incident Investigation


On January 27, 2009, almost a year to the day prior to the incident, a leak developed in the Oleum Tower

circulation piping. Although the amount estimated to have been released was greater than the January 23,

2010, release (40 pounds vs. 22 pounds), supervisors deemed the situation unnecessary for an emergency

shutdown and activation of a fume alert.

The emergency response for the 2009 incident was inconsistent with that taken in 2010. Unlike in 2010,

in 2009 a hot line 28 announcement informed plant personnel of the incident. In the incident

27
A mil is a unit of measure equal to one-thousandth of an inch (i.e., 1/1000 in).
28
A hot line announcement involves notification to a pre-determined list of operating and supervisory personnel
who are all informed of an incident at the facility with one call.

46
E.I. DuPont de Nemours & Co. Final Report September 2011

investigation report for the 2010 incident, no criteria is discussed that would provide guidance for the

appropriate response or what distinguished the two events.

3.3.4 PM Program Recommendation from 2009 Incident


The internal DuPont investigation identified the following key factor in the 2009 incident: Pipe in acid

service tends to have very localized areas of erosion/corrosion that can be easily missed while performing

thickness checks. These areas are often the result of welds, the heat affected area of welds, and,

disruptions or turbulence in the acid flow.

Although DuPont realized that certain wall thinning in acid service could go undetected, one

recommendation from this investigation was to incorporate all piping in oleum service into a PM

schedule; however, this recommendation was not completed prior to the January 2010 incident.

Moreover, the sample line involved in the January 2010 incident was not included in the PM schedule. An

interview with one of the engineers responsible for arranging for this equipment to be included in the PM

schedule revealed that the oversight occurred due to poor communication between DuPont and the

contractors hired to perform the PM inspections.

3.3.5 Mechanical Integrity


The piping material, 304L stainless steel, is acceptable to carry this concentration of oleum. The expected

rate of wall thinning would project the lifetime of the pipe to be approximately 40 years, and this pipe had

been in service for only 19. While the oleum sample line was within the design specifications, DuPont did

not address the corrosion issues associated with acid service.

47
E.I. DuPont de Nemours & Co. Final Report September 2011

3.3.6 Heat Tracing Design


The oleum sample line was heat-traced 29 with a steam tracing line comprised of -inch copper tubing

strapped to the outside of the sample line. The steam in the copper tracing line heats the sample line to

prevent the oleum inside from freezing. Steam tracing, however, can create hot spots and often does not

distribute heat evenly throughout its length. A preferred method is electric tracing, which can be easily

controlled and prevents hot spots through even heat distribution (Dillon, 1997).

As described in the Analysis Section, steam tracing played a significant role in the failure of the sample

piping. Once the oleum escaped containment, the copper tracing corroded away. The oleum and steam

then mixed, and the resulting extremely corrosive sulfuric acid created the larger hole. If an electric

tracing line had been used, as DuPont suggests for these conditions, the larger hole would not have

formed, reducing the magnitude of this incident.

3.4 Key Findings


1. An internal DuPont investigation report from a prior oleum leak recommended including all

piping in a PM thickness monitoring program. The CSB found no evidence that the piping in the

January 23, 2010, incident was included in the program.

2. The general wall thinning rate estimate for the oleum service was conservative. However, highly

localized corrosion attack cannot be predicted by this method.

3. Corrosion caused a small leak in the oleum pipe under the insulation.

29
The protection of a liquid-filled pipe against freezing by installing heat tubing or heating cable around or along the
pipe

48
E.I. DuPont de Nemours & Co. Final Report September 2011

3.5 Root Causes


1. DuPont did not adhere to industry recommended practices to use electrical tracing instead of

steam tracing.

2. A defect in the piping, undetectable by routine NDE techniques, allowed for a loss of

containment.

4.0 Phosgene Release (January 23, 2010)

4.1 Background
4.1.1 Phosgene
Phosgene, in liquid and gaseous forms, is colorless and highly toxic and has a characteristic odor of

freshly cut hay or grass, with a boiling point of 8 C (47 F), and is liquid in cold weather, gas in warmer

weather. At room temperature phosgene is a dense gas that is heavier than air. Phosgene is manufactured

through the reaction of carbon monoxide and chlorine and is used widely in industry as a chemical

intermediate for isocyanate-based insecticides, polymers, and pharmaceuticals.

Inhalation is the primary route of exposure to phosgene. The OSHA 8-hour TWA PEL for phosgene is 0.1

ppm 30; the NIOSH IDLH concentration is 2 ppm. The odor threshold 31 ranges between 0.4 and 1.0 ppm,

which is higher than the OSHA PEL; therefore, odor is not a reliable detection method for phosgene, as

injury may occur before the odor becomes prominent. Phosgene gas may irritate skin and eyes upon

contact at lower concentrations. Liquid phosgene contact with skin can also cause severe chemical burns

at higher doses.

30
The NIOSH- and ACGIH-recommended TWA concentrations are also 0.1 ppm for phosgene.
31
An odor threshold is the lowest airborne concentration that can be detected by a population of individuals. The
range of detection varies among individuals.

49
E.I. DuPont de Nemours & Co. Final Report September 2011

Phosgene inhalation can result in two mechanisms of injury to the respiratory tract, both of which can

result in pulmonary edema 32 at high concentrations. Inhaled phosgene slowly undergoes hydrolysis and

forms HCl, which results in upper respiratory irritation and burning sensations, cough, and chest

oppressions. Symptoms may not appear until several hours after exposure. Phosgene also reacts with

proteins in the pulmonary bronchioles and alveoli, disrupting the blood-air barrier in the lungs and

resulting in increased lung fluid. Pulmonary edema can be present in victims as long as 40 hours after

exposure and may last days depending on the concentration and duration of the exposure.

4.1.2 Phosgene Stainless Steel Hose Transfer Operation


The SLM unit runs on a campaign 33 basis and is divided into two processes: the front end and back

end. The front end process makes five isocyanate intermediate products. Phosgene used to produce the

five intermediate products is fed to a process from 1-ton cylinders stored in the phosgene shed at the SLM

unit. The phosgene cylinder storage shed is a covered, partially walled structure where the phosgene

transfer and storage operations occur (Figure 12). All equipment used for these purposes is in or around

the shed. The shed contains no mechanical ventilation or exhaust systems to control phosgene leaks, only

natural ventilation flowing through the shed wall opening from the atmosphere.

32
Pulmonary edema, which occurs when fluid accumulates in the lungs, leads to impaired gas exchange and may
cause respiratory failure. It is due to either failure of the heart to remove fluid from the lung circulation
("cardiogenic pulmonary edema") or direct injury to the lung parenchyma ("noncardiogenic pulmonary edema").
33
The front end of the SLM unit manufactures several types of isocyanate intermediates on a demand-based
schedule.

50
E.I. DuPont de Nemours & Co. Final Report September 2011

Figure 12. Phosgene shed and full (F) and empty (MT) cylinder locations on day of incident (not to scale)

During normal operation, two cylinders are staged on weigh scales and each is connected to the process

with two 0.25-inch diameter by 48-inch long PTFE-lined, 304 stainless steel overbraid hoses. One hose

transfers liquid phosgene to a steam vaporizer and one provides 70-psig nitrogen to the cylinder. The

scales record the weight of the in-service cylinder and when the container is nearly empty, an alarm

notifies the board operator, who then directs operators to switch to a full cylinder. This switch is

completed by opening valves to the full cylinder and closing valves to the empty cylinder. The hoses

remain coupled in this operation, and plant SOPs do not require enhanced PPE such as a fully

encapsulated suit and breathing air. Under normal operating conditions, the process consumes two to

three cylinders of phosgene per day.

51
E.I. DuPont de Nemours & Co. Final Report September 2011

The SOPs do require operators to don a fully-encapsulated suit with supplied breathing air when they

replace an empty cylinder with a full cylinder. After clearing all phosgene from the stainless steel hose

with a nitrogen purge under vacuum to a scrubber, the hose is isolated from the vent piping and

disconnected from the empty cylinder. Operators then replace the empty cylinder on the scale with a full

cylinder and connect the stainless steel hose to the new cylinder.

Maintenance mechanics replace stainless steel hoses in phosgene service when a work order is generated

to change-out the hoses. The DuPont SOPs for the change-out frequency of the nitrogen and phosgene

hoses directs replacement every 30 days. 34

A number of manufacturers fabricate hose assemblies to DuPonts specifications for phosgene and

nitrogen hoses, which arrive pre-assembled and are stored in plastic bags in the maintenance shop. Prior

to connecting the hoses to the phosgene cylinders, the maintenance mechanics install valves on either end

of the hose. Hoses removed from service are decontaminated in a water bath and then disposed.

4.1.2.1 VanDeMark Chemical, Inc.


VanDeMark Chemical supplies phosgene to the Belle plant in 1-ton cylinders. VanDeMark, located in

Lockport, NY, is the only North American company that both produces and distributes phosgene. It

distributes phosgene and phosgene derivatives in 1-ton cylinders. Each VanDeMark cylinder is 87 percent

full and contains 2,000 pounds of phosgene. Each U.S. Department of Transportation-regulated cylinder

has two valves with a seal plug screwed in the outlet covered by a flanged and gasketed bonnet to protect

the valves and prevent leaks during transport. The Belle plant receives phosgene cylinders via truck that

are unloaded at the phosgene shed; empty cylinders are loaded onto the truck and returned to

VanDeMark.

34
DuPonts former maintenance management process directed that hoses be changed every 2 months.

52
E.I. DuPont de Nemours & Co. Final Report September 2011

4.1.2.2 Use of Personal Protective Equipment (PPE)


DuPont safety procedures include two levels of PPE required for work in the phosgene cylinder shed on

the SLM unit, based on the connection status of the phosgene cylinders. When the phosgene cylinders are

connected to the process and no breaks in the phosgene lines are occurring, the standard required PPE for

the SLM unit is a hard hat, steel-toed safety shoes, safety glasses, flame resistant clothing (FRC), and a

phosgene indicator badge. Work with this level of protection includes

entering the phosgene shed to check cylinder scale weights,


opening and closing valves to switch from one cylinder to another, and
operating the crane when loading and unloading full or empty cylinders in the phosgene shed

The Belle Plant SOPs for disconnecting a phosgene cylinder require operators to wear a chemical suit

(gloves, boots, and hood) with supplied breathing air in addition to the PPE listed above while performing

the work. During all phosgene cylinder line break operations, another operator, wearing standard PPE,

stands outside the shed to monitor the breathing air supply of the operator performing the work.

At the time of the incident, the employee fatally exposed to phosgene was wearing the standard PPE. This

met DuPont operating standards for the task he was performing, because he was likely checking cylinder

weights in preparation for switching to the partially filled riverside cylinder. The Belle Plant PPE

requirements and SLM unit procedures did not require him to don a chemical suit, with supplied air,

during this activity.

4.1.2.3 Phosgene Indicator Badge


Belle Plant safety procedures require all personnel (operators, contractors, managers) and visitors in the

SLM unit to sign a log sheet and obtain a phosgene indicator badge from the SLM control room prior to

entry and to wear a phosgene indicator badge in their breathing zone (Figure 13). Phosgene indicator

53
E.I. DuPont de Nemours & Co. Final Report September 2011

badges change color when exposed to phosgene, and the color indicates the concentration 1 minute after

exposure. After 2 consecutive days of use, personnel using badges must discard and replace their

indicator badge to ensure accurate sensitivity. 35

Two types of phosgene indicator badges are available for use in the SLM unit. For work tasks not

involving supplied air, personnel clip SafeAir System phosgene badges (Morphix Technologies) to the

collar or pocket of FRC near the breathing zone. The badges change from white to pink or red to indicate

dose, concentration, or duration of exposure. In addition to badges, the SafeAir system uses a color

comparator wheel to detect exposure dose and the presence of phosgene between 0.9 and 100 ppm-min. 36

35
The manufacturing specifications state that the maximum recommended sampling time for each badge is 3 days.
The Belle plant requires phosgene badges to be replaced after 2 days to ensure accurate detection and avoid
discoloration or interference with other chemicals.
36
Parts per million-minute (ppm-min) is the concentration of a contaminant in air related to the exposure time
through inhalation; 48 ppm-min = 480 minutes of exposure at 0.1 ppm concentration.

54
E.I. DuPont de Nemours & Co. Final Report September 2011

Figure 13. SafeAir Phosgene Dosimeter Badge 37

For work tasks in the SLM unit requiring supplied air, all personnel must wear a CheckAir phosgene

badge inside the mask of their supplied air respirator. The CheckAir detector (Morphix Technologies)

detects exposure dose concentrations between 0.9 and 100 ppm-min. The color comparator wheel for

detecting exposure concentrations of the CheckAir detectors differs from that of the SafeAir badges.

4.1.2.4 Alarms
The SLM unit has 12 phosgene sensors placed in and around it to continuously sample and record

phosgene concentrations every 30 seconds; concentrations of phosgene are detected via an

electrochemical diffusion sensor within a range of 0.05 to 1 ppm. One phosgene sensor is located in the

phosgene shed, six are in the SLM building, and two are located outside the building. Three sensors are

37
The badge in Figure12 has a range of 0.5 to 450 ppm-min. The SafeAir badge worn by the exposed employee had
a range of 0.9 to 100 ppm-min.

55
E.I. DuPont de Nemours & Co. Final Report September 2011

on the fence line of the facility along the Kanawha River, approximately 120 feet from the phosgene

storage shed.

The analyzer readings are monitored by the DCS in the SLM control room, and concentrations in excess

of 0.05 set off audible and visual alarms at the board operators work stations. Concentrations equal to or

greater than 0.05 ppm set off a medium-high alarm and concentrations at or above 0.1 ppm set off a high-

high alarm. The CSB could find no evidence that audible or visual alarms were in service in the phosgene

shed when the release occurred.

On the day of the incident, the phosgene release activated alarms in the control room for four of the 12

analyzers in and around the SLM unit. The phosgene analyzer in the shed recorded concentrations ranging

from 0.04 to 1.0 ppm for approximately 50 minutes following the initial release. Two of the three fence

line monitors triggered alarms, with the maximum recorded concentration of 0.27 ppm on a monitor

located approximately 120 feet from the phosgene shed along the river. Another monitor, located on a

spill tank outside the SLM unit building, also recorded a concentration of 0.04 around the time of the

release.

All 12 phosgene analyzers have a maximum detectable concentration of 1 ppm. The analyzers do not

record actual values for concentrations in excess of 1 ppm; therefore, if phosgene concentrations exceed

the detection range at the analyzer sample point, the values are recorded only as 1 ppm.

4.1.3 Phosgene Highly Toxic Material Guardian Committee


DuPonts Phosgene Highly Toxic Material Guardian Committee focuses on the safe management of

phosgene at applicable DuPont facilities. DuPont has several guardian committees for highly toxic

materials (HTMs) used within the company. The committee is comprised of representatives, known as

phosgene guardians, from all DuPont sites that produce or consume phosgene. Managers from affected

processes, corporate health and safety representatives, engineers, and industrial hygiene specialists also

56
E.I. DuPont de Nemours & Co. Final Report September 2011

participate. The Phosgene Guardian Committee holds meetings twice a year to share learnings and discuss

phosgene handling issues.

DuPont has an HTM manual for phosgene, a company protocol that includes requirements and guidelines

for the safe design and operation of processes that generate or use phosgene. The primary purpose of the

manual is to reduce the likelihood of phosgene harming employees or the public. The requirements of the

manual are mandatory for all DuPont facilities with enough phosgene to impose a significant offsite

hazard as determined by a chemical consequence analysis of offsite exposure. The Phosgene Committee

conducts a second-party audit of all facilities using phosgene against the requirements and guidelines set

forth in the phosgene HTM manual approximately every 3 years. The Phosgene HTM Committee audited

the SLM unit at the Belle Plant in September 2006; the next audit was scheduled for January 25, 2010,

just two days after the phosgene release incident.

4.2 Incident Description


The third incident occurred on January 23, 2010, between 1:45 and 2:00 p.m. A stainless steel braided

transfer hose connected to a partially filled, but not in service 1-ton phosgene cylinder failed

catastrophically in the SLM unit phosgene shed. This incident occurred in the phosgene shed. When the

release occurred, an operator was in the phosgene shed inspecting the status of the riverside 38 phosgene

cylinder as he anticipated that the active cylinder was nearly empty and would need to be switched. He

was sprayed across the chest and face with liquid phosgene remaining in the riverside hose from a

previous transfer operation.

38
The cylinders are commonly referred to as hillside or riverside based on their orientation in the phosgene shed
relative to the hills north of the building and the Kanawha River to the south.

57
E.I. DuPont de Nemours & Co. Final Report September 2011

DuPont estimates that about 2 pounds of phosgene were released to the atmosphere when the hose failed.

The CSB concurs with this estimate and further calculated that the operator would have received a lethal

dose of phosgene in less than one-tenth of a second (Appendix D).

Immediately after the operator was sprayed, he called for assistance on the SLM unit public address

phone in the phosgene shed. A coworker who responded to the call noticed that the victims phosgene

dosimeter badge (Figure 13) was discolored, indicating an exposure. The coworker directed the exposed

worker to a plant truck to transport him to the plants medical center for assessment and treatment. As

they drove to the medical center, the two workers were met by the Shift Supervisor and the exposed

worker was transferred to the shift supervisors vehicle to complete the trip. While en route to the plants

medical center, the front gate guard was radioed and advised to call Metro 9-1-1 and request that an

ambulance respond for a medical emergency. The exposed worker, while at the medical center waiting for

the ambulance, chose to wash his face and hands, but there is no evidence or record that he was placed in

a safety shower to wash off, as instructed by the emergency procedures, or that any decontamination

activity took place beyond the hand and face washing. He was given a change of coveralls to put on in

exchange for the work clothes he was wearing. The gate guard called Metro 9-1-1 at 1:59 p.m., requesting

transport for a medical emergency patient to the hospital. The 9-1-1 dispatcher asked if there was a

chemical release; however, the gate guard, who was unaware of the situation, responded that there was no

release and that the response was for a medical emergency. As part of the Metro 9-1-1 emergency

response protocol, the dispatcher asks for specific information to ensure that responders are as informed

as possible prior to arrival at the scene. At 2:03 p.m., an ambulance was dispatched from the KCEAA.

At 2:08 p.m., responding EMTs asked Metro dispatchers if more information was available about the

victim. When Metro called DuPont to get more information, the line was busy. EMTs also wanted to

know if there was a chemical exposure, but Metro 9-1-1 could not get that information from DuPont. Six

minutes later, the EMTs arrived at the DuPont gates.

58
E.I. DuPont de Nemours & Co. Final Report September 2011

EMTs were directed to the DuPont medical center to meet the exposed worker. As the EMTs gathered the

worker for transport, they were given a written phosgene treatment protocol intended to be used at the

hospital to provide treatment. While the worker was being transferred to their care, DuPont employees

told the EMTs that the victim had been exposed to liquid phosgene.

The EMTs left the facility with the victim at 2:26 p.m., or 27 minutes after the first call to Metro 9-1-1.

During transit and after arrival at the hospital at 2:34 p.m., the victim was lucid, conscious, and talking

clearly to the emergency responders and attending physician. Until the attending ER physician consulted

the company-provided phosgene treatment protocol, which advised 48-hour monitoring for suspected

phosgene exposures, he considered sending the victim home based on his condition shortly after arriving

at the hospital. A baseline X-ray revealed no congestion in the victims lungs. At about 5:30 p.m., or

almost 4 hours after exposure, the operators condition began to rapidly deteriorate. Over the next 29

hours, the victim received treatment from a variety of physicians, but his condition failed to improve and

he died at 9:27 p.m. on Sunday, January 24, 2010.

Post-incident, KCEAA staff voiced concerns regarding the quality and timeliness of information DuPont

provided to Metro 9-1-1 dispatchers and responding EMTs. The concerns raised address the need to

ensure that emergency responders and their equipment are not exposed to contaminants and that the

victims they are assisting receive optimum care in transit for medical treatment. A review of comparable

responses by KCEAA EMTs in the region reveal that the response time to DuPont and from there to the

hospital was not unduly delayed by the lack of information. A sampling of similar emergency responses

reveal an average response time from the initial call to Metro 9-1-1 until arrival at the hospital to be about

36 minutes. Total elapsed time for the response time on the day of the exposure was 35 minutes.

Although the emergency response and transport of the victim was not delayed during this incident or the

oleum release, because of a lack of clear, accurate information regarding the material involved, response

procedures have since been modified by Metro 9-1-1 administrators. These modifications mandate that

59
E.I. DuPont de Nemours & Co. Final Report September 2011

EMS units not report directly to the site of an incident until clear information has been provided such that

EMS personnel will not be at risk of unknown contaminants/threats. This change in response protocol

was incorporated after several incidents in the Kanawha Valley. The CSB considers the change in

response protocol significant enough to define the cause and effect of the communication gap as a near-

miss. Several key factors that contributed to poor communication, including the absence of a process

knowledgeable person assigned to convey information to the dispatchers and the lack of a direct line to

the Metro 9-1-1 emergency operations center, must be recognized and addressed.

One confirmed and one possible phosgene exposure to workers occurred after the initial release. The first

was when a coworker responded to the call for assistance immediately after the phosgene hose ruptured.

As he drove the victim to the facilitys medical building, the coworkers dosimeter badge became slightly

discolored, indicating phosgene exposure.

A possible source of this exposure was phosgene vapor in the atmosphere as recorded on one of three

fence line monitors about 120 feet from the shed along the river. Another possible source was the victims

clothing, which may have been saturated with phosgene immediately after the release. When interviewed,

this employee said that pulmonary function tests performed afterward showed no signs of adverse effects.

A second possible exposure occurred when an employee working in the SLM unit went toward the

phosgene shed shortly after the release. He reported in an interview that as he got closer, he noticed a

smell that he had not encountered before or since. He recalled that the odor was not strong or offensive as

would be expected with ammonia or chlorine, but noticeably different from any odors he had smelled in

the past. Being unfamiliar with the characteristic fresh mown hay odor associated with phosgene, he left

the area.

Although the phosgene shed area has flashing lights to alert against entry into the area during cylinder

changes, there is no evidence that a fume, medical, or plant radio alert sounded at any time during this

release episode to warn operators and maintenance personnel to avoid coming near the phosgene shed.

60
E.I. DuPont de Nemours & Co. Final Report September 2011

4.2.1 Community Impact


Two of the three fence line analyzers recorded a maximum concentration of 0.15 and 0.27 ppm 39

phosgene, indicating that phosgene concentrations had traveled offsite toward the Kanawha River.

However, no member of the public reported phosgene exposure symptoms the day of the incident nor did

the U.S. Coast Guard restrict river traffic or conduct air monitoring as it had a day prior as a result of the

methyl chloride release.

4.3 Incident Analysis


4.3.1 Hose Failure Analysis
Post-incident inspections of the stainless steel hoses used for the two phosgene cylinders connected to the

process identified comparable degradation patterns. Their failure was associated with corrosion that

developed in approximately the same location on hoses used to transfer phosgene from the riverside and

hillside cylinders.

Investigators found that while the majority of tags attached to the hoses to indicate the intended service

were secured in place with plastic ties and metal clampsas was normalone manufacturers tag was

secured with white plastic adhesive tape (this tag applied by the manufacturer also provided identification

information). The corrosion identified on the two hoses associated with the hillside and riverside

cylinders was localized under the area covered by the white plastic adhesive tape securing the tag. 40 The

characteristics of the transfer hose, consisting of a core constructed of permeable PTFE and braided 304-

39
ERPG-2 value for phosgene is 0.20 ppm and at this concentration all could be exposed for up to one hour
without experiencing or developing irreversible or other serious health effects or symptoms that could impair their
abilities to take protective action (AIHA, 2008).
40
Witnesses could not provide an exact date that the hoses came into the facility with the tags affixed with adhesive
tape.

61
E.I. DuPont de Nemours & Co. Final Report September 2011

stainless steel, provided a suitable environment under the adhesive tape for stress corrosion cracking

(SCC 41) to occur.

To provide comparative data, hoses from the hillside, riverside, and exemplars of similar age and new

assembly were sent to an analytical lab for testing and analysis. The tests established that all of the hoses

were constructed with 304-stainless steel and the construction material for the inner core of the hoses was

PTFE, as expected.

4.3.2 Effect of Plastic Adhesive Tape


The PTFE, 304 stainless steel, and white plastic adhesive tape contributed to the incident. The PTFE inner

core was permeable and susceptible to phosgene vapor diffusing through the hose. The adhesive tape used

to secure the tag contributed to the retention of phosgene gas on the exterior of the stainless steel

overbraid. The phosgene gas converted to HCl, and 304-stainless steel overbraid is subject to corrosive

attack by HCl. Since the white plastic adhesive tag trapped the phosgene permeating through the PTFE

inner core, the resulting concentration of HCl was much higher under the tag than elsewhere on the hose

(Figure 14).

41
Stress corrosion cracking is the formation of brittle cracks in a normally sound material through the simultaneous
action of a tensile stress and a corrosive environment.

62
E.I. DuPont de Nemours & Co. Final Report September 2011

Figure 14. The distinct effect of the white plastic adhesive tag on the corrosion of the stainless steel
overbraid

Additionally, at the time of the incident, the isolation valves on the phosgene hose on the riverside

cylinder were closed, which retained liquid phosgene in the hose and pipe between the valves that isolated

the cylinder from the process. The heavy corrosion of the stainless steel overbraid, coupled with the time

the hose had been in service and thermal expansion 42 of the isolated liquid phosgene, caused the hose to

fail catastrophically. When this failure occurred, the worker was exposed as he walked nearby to check on

the status of the adjacent in-service cylinder.

42
Tendency for solids, liquids and gases to change in volume in response to a change in temperature.

63
E.I. DuPont de Nemours & Co. Final Report September 2011

4.3.3 Hose Degradation Issues


Although the maintenance plan for the hillside and riverside hoses prescribed a regular change-out

schedule of 30 days, work orders show that change-out frequency was neither systematic nor predictable.

At least three times from 2006 to 2010, phosgene hoses were left in service from 4 to 7 months.

4.3.4 Hose Change-out Frequency


Several times each year, the phosgene process is halted so the plant can produce a material requiring the

physical removal of phosgene, including all full or empty 1-ton cylinders, from the phosgene shed.

Table 2 shows the change-out frequency of the phosgene hoses in the SLM unit and the periods when

SLM did not run processes using phosgene. The most recent recorded instance where phosgene was not

used in the process was between September and November 2009, 2 months prior to the incident. Work

orders for changing-out the phosgene hoses indicate that the stainless steel transfer hoses connected at the

time of the incident had been in service for more than 6 months. This included a removal of the phosgene

system change-out in September 2009 when the hoses could have been changed-out.

64
E.I. DuPont de Nemours & Co. Final Report September 2011

Hose Change-out Frequency


Month/Year Phosgene Hoses Phosgene Used
Jul-05 Changed Phosgene Used
Aug-05 Changed
Sep-05 Changed
Oct-05 Changed
Nov-05 Changed
Dec-05
Jan-06 Phosgene Used
Feb-06 Phosgene Used
Mar-06 Changed Phosgene Used
Apr-06 Phosgene Used
May-06 Phosgene Used
Jun-06 Changed Phosgene Used
Jul-06 Changed Phosgene Used
Aug-06
Sep-06
Oct-06 Changed
Nov-06 Changed
Dec-06
Jan-07 Phosgene Used
Feb-07 Phosgene Used
Mar-07 Changed Phosgene Used
Apr-07 Phosgene Used
May-07 Changed Phosgene Used
Jun-07 Phosgene Used
Jul-07 Phosgene Used
Aug-07
Sep-07 Changed
Oct-07
Nov-07
Dec-07
Jan-08 Phosgene Used
Feb-08 Phosgene Used
Mar-08 Phosgene Used
Apr-08 Changed Phosgene Used
May-08 Changed Phosgene Used
Jun-08 Phosgene Used
Jul-08 Changed Phosgene Used
Aug-08 Changed
Sep-08
Oct-08
Nov-08
Dec-08
Jan-09 Changed Phosgene Used
Feb-09 Phosgene Used
Mar-09 Phosgene Used
Apr-09 Phosgene Used
May-09 Phosgene Used
Jun-09 Changed Phosgene Used
Jul-09 Phosgene Used
Aug-09
Sep-09
Oct-09
Nov-09
Dec-09
Jan-10 Phosgene Used

Table 2. Phosgene hose change-out frequency

65
E.I. DuPont de Nemours & Co. Final Report September 2011

The CSB found that change-out frequency was intended to be governed automatically by the Belle

facilitys SAP maintenance program. Some supervisors also relied on the maintenance coordinator

remembering to initiate the change-out.

4.3.5 SAP Work Process


DuPont uses the plant maintenance module of SAP enterprise resource planning software 43 to schedule

the change-out of phosgene hoses at pre-determined 30 day intervals. The SAP system is programmed to

issue the work orders for hose replacement to prevent the release of phosgene; thus, maintaining accurate

data in the SAP database is crucial to protect against phosgene exposure (Appendix C).

In late 2006, SAP data managing the change-out frequency of the phosgene hoses at the Belle facility

were changed; consequently, SAP stopped automatically issuing work orders to change the hoses, but

plant personnel were unaware that SAP no longer automatically issued the work orders. The CSB

requested additional information regarding the change; however, DuPont could not determine who

changed the SAP data, why it was changed, or when the change was executed. No back-up layer of

protection, such as a weekly critical equipment maintenance check sheet or an inspection tag, ensured that

the hoses were changed at the pre-determined frequency. With SAP no longer automatically issuing work

orders to change the hoses, the system did not trigger maintenance notifications to change-out the hoses at

assigned intervals.

43
Enterprise resource planning software is a type of database that allows data related to flows of money and other
resources in areas such as accounting, supply chain management, sales and marketing, manufacturing,
maintenance, and project management to be recorded and accessed.

66
E.I. DuPont de Nemours & Co. Final Report September 2011

4.3.6 Near-Miss Phosgene Incident


On the morning of the phosgene incident, operators asked maintenance personnel to replace the phosgene

hose on the hillside cylinder because of a suspected flow restriction. Although the cylinder was still about

half full, it was removed from service and replaced with the full riverside cylinder.

The hillside phosgene supply hose and valve assembly were removed and decontaminated in a water bath.

When the hose was removed from the water, the white adhesive ID tag had fallen off, revealing a broken

stainless steel braid and collapsed PTFE liner, a possible cause of the flow restriction (Figure 15).

Figure 15. Damaged hillside phosgene hose removed from phosgene cylinder. The plastic adhesive tag
that covered the damaged section fell off during the hose decontamination procedure.

An operator stated during an interview that when he saw the physically defective section of the frayed

hose, he told his coworkers, stressing that the hose was close to rupturing and that they were lucky to

have found it and changed-out the hose in time. Unfortunately, this discovery was not captured as a near-

miss, since supervisors were not made aware of the issue.

67
E.I. DuPont de Nemours & Co. Final Report September 2011

Operators told the CSB investigators that they had never seen a phosgene stainless steel hose braid

corroded to the point of separation. Although they were surprised and concerned about their finding, and

since supervisory staff does not work on weekends, they planned to tell the supervisors about the

discovery on Monday morning, about 48 hours later. Operators said that they expected that the

supervisors would conduct a full investigation; however, since the incident occurred on a Saturday, it was

not investigated. Had there been a system in place for operators to report near-miss incidents on

weekends, the near-miss investigation may have been properly initiated prior to the fatal release.

4.3.7 Mechanical Integrity


The DuPont P3H Standard lists acceptable construction materials for flexible hoses used in HTM service

and recommends three different hoses acceptable for use with phosgene: H2, H7, and H9 (Table 3).

68
E.I. DuPont de Nemours & Co. Final Report September 2011

DuPont P3H Standard Hoses for Phosgene Service


Name Specifications

H2 Inner core material: Monel 400, corrugated


Reinforcement material: Monel 400 overbraid
End fitting material: Monel 400 SCH. 80
Core/fitting connection method: Welded, full penetration

H7 Inner core material: Hastelloy C276, corrugated


Reinforcement material: Monel 400 or Hastelloy C276 overbraid
End fitting material: Hastelloy C276 stub ends
Core/fitting connection method: Welded, full penetration

H9 Inner core material: Teflon 44 PTFE, helical, corrugated, taped or extruded


construction, unpigmented or conductive
Reinforcement material: PVDF (Kynar) double overbraid
Monel 400, Hastelloy C276, or Teflon encapsulated
End fitting material: SS
Core/fitting connection method: Crimped (or swaged)

Table 3. Flexible hoses for phosgene service as listed in the DuPont P3H Standard: Flexible Chemical
Hose for Highly Toxic Services

The Belle facility did not use any of the P3H specified hoses and configurations; instead, it used a flexible

hose made of a Teflon PTFE inner core and a braided stainless steel reinforcement material, even though

stainless steel is not recommended for phosgene service, as it is susceptible to SCC from chlorides.

Phosgene, which can readily react with air to produce chlorides, can permeate PTFE, directly exposing

the stainless steel braid to chloride attack.

44
Teflon is the DuPont-registered trademark for PTFE.

69
E.I. DuPont de Nemours & Co. Final Report September 2011

4.3.8 Flex Hose Materials of Construction


The Belle facility referred to corporate experts and the La Porte 45 facility, where flexible hoses were

being used for phosgene service.

The discussions between the two plants and corporate experts about flexible hoses began in 1987, when

corporate experts suggested the use of Monel metal for both the hose core and hose overbraid, since it

resists chloride SCC. However, the La Porte plant asserted that its history with Monel metal was less than

desirable; one correspondent noted, The La Porte plant was considering testing Kynar overbraid-
46
covered Teflon hose because of discoloration and gradual deterioration of the Monel.

An expert from DuPont corporate told Belle that the discoloration was not a problem:

Reports from La Porte that Monel braided hoses were corroding in phosgene

service are not exactly true. The hoses at that time were Teflon lined, with a

Monel outer overbraid. Due to permeation of phosgene through Teflon, the

Monel was slightly attacked, forming a green surface film known as a patina 47

which is common to all copper-based alloys.

A Belle representative sent a questionnaire to La Porte in August 1987 to evaluate its hose program. The

questionnaire revealed that La Porte had been using PTFE-lined stainless steel hoses for the previous 3 to

4 years and that they were replaced every 3 months. It reported that the majority of the hose failures were

due to fatigue, and that the facility was using stainless steel because it is not as susceptible to failure from

45
DuPont uses phosgene at four of its facilities. DuPont no longer uses phosgene in La Porte, TX.
46
The Kynar hose was also not pursued due to pre-conceived flexibility limitations.
47
Patina, most easily observed on old pennies, is a green film formed naturally on the surface of copper and copper-
based metals.

70
E.I. DuPont de Nemours & Co. Final Report September 2011

fatigue and bending stresses as are Monel and Kynar hoses. After reading the questionnaire, the corporate

DuPont expert wrote,

I still believe that Monel is the best choice for material of construction for

phosgene unloading hoses (and definitely for the fittings). I am surprised that La

Porte is using Teflon-lined hose with stainless overbraid since Teflon is known to

be permeable and the phosgene is known to attack the stainless.

The DuPont expert further stated,

Admittedly, the Monel hose will cost more than its stainless counterpart.

However, with proper construction, and design so that stresses are

minimizeduseful life should be much greater than 3 months. Costs will be less

in the long run and safety will also be improved.

Correspondence or other records that would explain why the experts recommendation went unheeded at

La Porte and why the Belle staff decided to follow the La Porte approach was not discovered during the

CSB investigation. However, Belle decided to follow La Portes example, and adopted a hose design not

recommended by its P3H Standard or by a DuPont corporate expert.

The phosgene hose replacement frequency at Belle is defined in DuPonts Phosgene Hose Assembly

Procedure: Due to the extremely hazardous nature of phosgene the hose assemblies are replaced every 2

months.

However, the PM schedule in SAP is actually set to a replacement frequency of 30 days. This procedure

does not effectively communicate why the hoses must be replaced so frequently: if left on too long, the

accepted corrosion condition poses a serious risk to the facility and the community.

71
E.I. DuPont de Nemours & Co. Final Report September 2011

Figure 16. Flex hose comparison photographs: (top to bottom) ruptured riverside hose, flow restricted
hillside hose, a new hose with attached ID tag

4.3.8 Non-routine Job Planning


Operators told the CSB investigators about the difficulty maintaining the required flow of phosgene from

one of the two cylinders on the weigh scales the day prior to the exposure incident. The phosgene flow

from the cylinder to the process was inadequate; thus, they performed a non-routine operation to establish

a steady flow of phosgene because they suspected a plugged hose or a malfunctioning automatic feed

control valve. Non-routine operations are characterized by infrequent practice, can be both planned and

scheduled, or can occur without scheduling

To minimize disruption of the phosgene flow to the process, operators switched to the riverside cylinder,

which operated as expected and supplied the normal flow rate. Continuing throughout the day and into the

next, operators repeated switching to the riverside cylinder as the flow from the hillside cylinder became

low enough to begin to affect the process. When valves for each of the respective transfer hoses were

72
E.I. DuPont de Nemours & Co. Final Report September 2011

closed, liquid phosgene was not evacuated as required by the SOP for switching from one cylinder to

another. Since the operators were not fully aware of the hazards of thermal expansion, liquid phosgene

remained in the hoses as the cylinders were switched.

The CSB investigators reviewed DCS flow and weight data and saw a distinct difference in the ability of

the riverside cylinder to provide the needed flow rate of phosgene compared to the hillside cylinder in this

operation. All DCS information the operators received as a result of the non-routine cylinder switching

indicated that their actions were successfully maintaining the smooth operation of the unit.

The operators, however, were involved in non-routine operations by attempting to maintain steady-state

operations, as the SOPs did not address handling flow restriction. In addition, they were unaware of the

threat of liquid thermal expansion developing as a result of switching the cylinders and not evacuating the

hoses after each switch-out operation.

4.4 Process Hazard Analysis


PHAs were conducted on the phosgene cylinder feed system and vaporizer as part of the Front End SLM

Unit assessment in 1994, 1999, 2004, and 2009. The 2009 PHA team, all DuPont employees, included a

senior process engineer, two technical resources, a mechanic, and a front end operator; reviewed subtle

changes to the process and associated MOC documentation since the last PHA in 2004 and previous

phosgene release incidents, and recommended corrective actions. The PHA for the phosgene system

included the 1-ton cylinders, nitrogen pressuring system, the vaporizer, and all associated piping and

controls. The team used a Hazard and Operability 48 (HAZOP) and What If 49 methodologies to review

process hazards and deviations.

48
A systematic method in which process hazards and potential operating problems are identified using a series of
guidewords to investigate process deviations (CCPS, 2008).

73
E.I. DuPont de Nemours & Co. Final Report September 2011

The team recognized and assessed the potential for a phosgene release from the cylinder transfer hoses

but only if the hoses were incorrectly connected or inadvertently disconnected while the cylinder feed

valve remained open. They did not assess the potential for the hose to rupture due to thermal expansion of

liquid phosgene even though the potential for liquid phosgene thermal expansion was evaluated in other

process equipment during the 2009 PHA.

None of the consequence scenarios the PHA team assessed involved failure of the phosgene transfer hose

or the nitrogen flex hose. When the team evaluated the phosgene vaporizer, it considered corrosion

potential when stainless steel is exposed to phosgene and water, but did not apply those factors to the

cylinder transfer hoses. For the vaporizer, the probability value assigned to the phosgene leak scenario

was decreased by reliance on the PM program to detect corrosion. The PHA team also noted that the

slowly developing corrosion would decrease the probability of a leak because the corrosion would be

noticeable during visual inspections. If the PHA team had assessed the thermal expansion and corrosion

issues for the phosgene transfer hoses and had applied the same conditions to decrease the probability as

used for the vaporizer corrosion scenario, the incident may still have occurred due to the teams reliance

on the PM program to reduce the hazard. Unfortunately, the slowly developing corrosion on the hose was

not visible due to the location of the white plastic adhesive tape, and the PM program was not configured

to ensure that the hoses were changed at the appropriate frequency.

Phosgene permeation through PTFE had resulted in leaks at Belle in the past; however, the PHA team did

not consider this hazard for the phosgene cylinder hoses. The CSB received documentation of all SLM

PHA audits dating back to 1994. The 1999 PHA included two incidents in which phosgene likely

permeated through PTFE-lined conveyance equipment in other parts of the phosgene process. Even with

49
A technique in which a team with process knowledge and experience examines possible process deviations or
combinations of deviations than can result in an undesired consequence (CCPS, 2008).

74
E.I. DuPont de Nemours & Co. Final Report September 2011

these previous incidents considered, the PHA team still did not account for the potential of the phosgene

cylinder hoses to result in a release under similar conditions.

4.5 Audits
4.5.1 Unit Second-Party PSM Audit
In August 2007, a second-party audit team of engineers and health and safety experts from other DuPont

facilities audited the SLM unit against regulatory and company PSM requirements. As in the F3455 unit

audit, the team focused on MOC-subtle change, pre-startup safety reviews (PSSRs), training, PHAs,

mechanical integrity, and process technology. The audit contained 64 findings27 observations, 35

policy, and two regulatory issueswithin the F3455 and SLM units at Belle.

One regulatory issue noted for the SLM and F3455 units was timely initiation of accident investigations.

Auditors noted several instances where incident investigations were not started and communicated within

the Belle plant 24- or the 48-hour OSHA requirements. The audit team recommended revising the Belle

Plant Incident Investigation procedure and area practices to ensure that plant personnel initiate

investigations within 24, and no later than 48, hours following an incident. According to the audit

tracking plan the CSB investigators reviewed, an assigned DuPont employee completed and closed the

recommendation as of June 2009.

However, in the case of the hillside hose near-miss prior to the phosgene exposure (Section 4.3.3),

operators told the CSB investigators that they planned to communicate the near-miss to supervisors for

investigation the following Monday; however, this would not have been within the Belle Plant required

24-hour period. The OSHA PSM Standard requires the employer to investigate each incident which

resulted in, or could reasonably have resulted in, a catastrophic release of highly hazardous chemical in

the workplace (1910.119(m)(1)) and that an incident investigation shall be initiated as promptly as

possible (1910.119(m)(2)). The EPA Risk Management Program also requires an investigation of an

incident involving a regulated substance, such as phosgene, be initiated within 48 hours (40 CFR part

75
E.I. DuPont de Nemours & Co. Final Report September 2011

68.81(b)). Though supervisors are not typically at the facility on weekends, management and safety and

health experts, including the SLM Area Manager, were at the Belle Plant the morning of Saturday,

January 23, 2010, attending the safety pause meeting. Had the incident been reported in a timely manner,

management onsite could have immediately initiated an investigation.

4.5.2 Onsite Phosgene Generation


In 1988, DuPont engineers considered two options for using phosgene at the Belle facility: in cylinders

from an offsite provider or constructing a phosgene generation plant to make phosgene onsite. To better

understand the hazards involved in each design, DuPont engineers conducted a risk assessment in which

four cases were considered (Table 4):

Case 1. Operating with a liquid phosgene feed from cylinders


Case 2. Vaporizing the feed from the cylinders
Case 3. Installing a plant to make phosgene from CO and Cl2
Case 4. Enclosing the phosgene plant (in a fully contained building with an air scrubber)

After evaluating each case, they estimated the risk of fatality as follows:

Onsite Fatalities per 10,000 years Offsite Fatalities per 10,000 years
Case 1 244 10.5
Case 2 154 0.22
Case 3 16.7 0.007
Case 4 2.3 0.006

Table 4. Preliminary risk assessment by DuPont Engineering, 1988 (Appendix E)

While Case 4 was estimated to have the least amount of risk, the assessment concluded,

Spending $2 MM for an enclosure to get from Case 3 to Case 4 saves 14.4 lives

per 10,000 years. (Almost all the improvement is in on-site risk. Off-site risk

improvement is not significant.) This sets a value of life plus public outrage at

$143 MM. It may be that in the present circumstances the business can afford $2

MM for an enclosure; however, in the long run can we afford to take such action

76
E.I. DuPont de Nemours & Co. Final Report September 2011

which has such a small impact on safety and yet sets a precedent for all highly

toxic material activities [?].

After the analysis, construction on Case 3, the open-to-atmosphere phosgene generation plant, began.

However, the phosgene generation plant was abandoned mid-construction, and Case 2 is the current

configuration at the Belle facility.

Documentation to support why the phosgene generation plant was abandoned was not provided, although

the CSB obtained a proposal by a third-party contractor to build the plant. The proposal for a plant, as

presented in Case 3, estimated a cost of $830,000 and stressed the contractors history of building

successful phosgene generation units. DuPont did not act on this proposal; anecdotal evidence from

interviews suggests that corporate engineers decided to use DuPont resources to construct the plant.

However, once the project was partially complete, the effort was abandoned as it was determined that the

DuPont-designed system would not work.

DuPont cancelled plans for the enclosed phosgene generation unit, but the potential for offsite impact still

remained a concern and was identified in SLM unit PHAs years later. In 2004, a PHA on the SLM unit

by Belle Plant personnel identified the need for a shed enclosure with a scrubber to mitigate or prevent the

release of phosgene offsite. The recommendation resulted from a What if analysis during the PHA.

The PHA team listed two separate scenarios that could result in a plant-wide or offsite consequence, both

recommending a shed enclosure. The original due date for the shed enclosure was scheduled for

December 2005 but extended to December 2006; three subsequent extensions on the enclosure

recommendation remained incomplete the day of the fatal phosgene release (Table 5).

77
E.I. DuPont de Nemours & Co. Final Report September 2011

PHA Enclosure Recommendation Delays


2004
Original Recommendation created in SLM 2004 PHA
"Provide appropriate mitigation to prevent multiple fatalities from the Dec-04
release of a 2000 lb phosgene cylinder."
Due Date: Dec-05
2005

First Extension
"A COC12 generation system is currently being
evaluated, and if this was installed the shed enclosure may be May-05
designed differently to handle the appropriate chemicals."
New Due Date: Dec-06
2006

Second Extension
"Work to define the scope on this item is progressing but not yet
complete. We are evaluating potential lower cost alternatives to total Dec-06
shed enclosure."
New Due Date: Dec-08
2008

Third Extension
"...the schedule indicates completion by August 2009."
"The holds on the capital project were due to uncertainty of the Dec-08
future of the facility and due to the cost of the project."
New Due Date: Nov-09
2009

Fourth Extension
" project to install a phosgene scrubber to address these
recommendations, an error in basic data was discovered. This Nov-09
invalidated the original design basis for the scrubbing system, and
required a halt to the project activity."
New Due Date: Nov-10
SLM 2009 PHA Completed
32 Recommendations are made, none of which capture the Dec-09
outstanding recommendation from the SLM 2004 PHA
2010
Fatal Phosgene Incident Occurs Jan-10

Table 5.Delay for Completing the PHA Recommendation for Enclosing the Shed

78
E.I. DuPont de Nemours & Co. Final Report September 2011

Following the phosgene release incident, DuPont announced that it would idle the storage and use of

phosgene at the Belle site for 2011 and later told the CSB that the site has permenantly discontinued all

onsite phosgene operations. The CSB requested documentation from DuPont that defines the status of the

PHA recommendation for the shed enclosure as of the date of this report. DuPont extended the PHA

recommendation for the shed enclosure until November 2010; however, the work was not completed and

was extended again until the end of 2011.

4.5.3 2006 Phosgene Committee Audit


In 2006, the Phosgene Guardian Committee audited against the DuPont Phosgene Highly Toxic Materials

(HTM) Manual, which included a review of the phosgene cylinder storage shed, the SLM production area,

and other areas of the Belle plant. Three audit team members from other DuPont sites visited the Belle

facility to conduct field walkthroughs and hold discussions with process unit personnel. The audit team

divided the findings and recommendations from the audit into two categories: policies and observations.

The policies were related to the requirements of the HTM manual and the observations were suggestions

or preferred, but not mandatory, practices.

The team found no regulatory compliance deficiencies in the audit, but did issue five policy

recommendations and eight observations. The policy recommendations applied to equipment downstream

of the phosgene cylinder feed system, including a recommendation to add inspection plans for corrosion

detection of the Teflon-lined reactor piping. The team found, and noted as an observation, that the hoses

used on the phosgene feed system were not one of the three types recommended for phosgene service by

the DuPont P3H Standard, but did not require the Belle facility to use the appropriate hoses.

The team also observed that liquid phosgene lines in the shed had moderate external corrosion and that

significant moisture in the shed should be addressed to eliminate future corrosion potential. Because these

items were observations, the HTM manual did not require that DuPont develop an action plan to resolve

79
E.I. DuPont de Nemours & Co. Final Report September 2011

them. Consequently, the Belle plant continued to use a hose for phosgene service that the company

standard did not recommend.

SLM unit equipment selection practices did not align with the requirements and recommendations in the

phosgene HTM manual. The manual states, Materials of construction must be selected properly to

handle phosgene safely but only recommends against the use of nonmetals for piping, valves, and

process equipment containing phosgene. It further states, Where small amounts of phosgene are present,

stainless steel lined with Teflon is commonly used without specifically quantifying an amount of

phosgene where Teflon is acceptable. In the SLM phosgene transfer system, phosgene was continuously

present in the PTFE-lined hoses while the connected cylinder was feeding the process.

The HTM manuals design information section requires that special attention be given to the prevention

of over pressuring those lines and vessels where liquid phosgene can be trapped between two isolation

valves. In the course of switching between cylinders on the morning of the phosgene incident, SLM

operators blocked in (i.e., closed the valve on each end of the hose), which trapped liquid phosgene

between the partially filled riverside cylinder and the valve to the process. The liquid phosgene trapped in

the hose underwent thermal expansion, rupturing the hose due to the overpressure of the line that was

facilitated by the weakened and corroded stainless steel overbraid. None of the SOPs for the SLM unit

warned against blocking in liquid phosgene to prevent hose ruptures, making operators less aware of the

thermal expansion hazards of phosgene.

4.6 Standards and Guidelines


4.6.1 DuPont Highly Toxic Materials Phosgene Manual
The DuPont HTM manual includes mandatory criteria for the storage, handling, maintenance, and

management of phosgene in quantities with the potential to cause offsite impact if released. The 86-page

manual also includes non-mandatory practices for new and existing units or facilities handling phosgene,

and company requirements and procedures related to first aid and medical treatment, MOC, design

80
E.I. DuPont de Nemours & Co. Final Report September 2011

information for new and existing phosgene equipment, and PSM principles. The Phosgene Guardian

Committee reviews and revises the manual and the committee chairperson and SHE leader authorize the

revisions. The Responsible Care Core Team reviews and approves all changes to mandatory requirements

before issuing the revised manual. The Plant or Unit Manager must authorize any deviation from the

manual requirements before using an alternative practice. The HTM Committee conducts a safety analysis

to ensure that the alternate practice is acceptable before implementation.

4.6.2 American Chemistry Council (ACC) Phosgene Safe Practice Guidelines


Manufacturers and users of phosgene formed the Phosgene Panel in 1972 to share information about

practices to safely produce, handle, and use phosgene throughout industry. The Phosgene Panel is part of

the Chemical Products and Technology Division of the ACC, an industry trade association for chemical

companies; its Chemical Products and Technology Division supports companies through continuous

evaluation and communication improvements related to the safe use of hazardous chemicals. Engineers,

health and safety experts, and occupational health physicians from member companies participate on the

panel, 50 which meets twice a year to share information and experiences related to handling phosgene. The

panel sponsors engineering studies and research to prevent phosgene-related incidents and has prepared

manuals for phosgene safe practices and medical treatment information as a resource for ACC member

companies.

The ACC Phosgene Panel compiles information from member companies into the Phosgene Safe Practice

Guidelines Manual to provide general information to those that manufacture or handle phosgene. The

manual contains nine sections of phosgene safety information such as phosgene properties, design

50
In 2010, all U.S. phosgene manufacturers participated in the panel: BASF Corp.; Bayer Corp.; Chemtura; Dow
Chemical; DuPont; Huntsman; SABIC Innovative Plastics; and VanDeMark Chemicals, Inc.

81
E.I. DuPont de Nemours & Co. Final Report September 2011

information for phosgene process facilities, transportation, emergency planning, first aid and medical

treatment, and training.

Phosgene panel members draft summaries of industry practices that they submit for review and approval

by all members of the ACC Phosgene Panel prior to inclusion in the manual. The panel periodically

updates the manual and adds new and relevant practices identified by industry. The ACC does not intend

for the manual to be a training tool or be adopted as procedure; it is to be referenced for general

information regarding safe practices for phosgene storage and use.

The Design of Facilities section of the manual has several subsections pertaining to construction

materials and layout of phosgene process equipment and facilities. This section includes leak prevention

information such as equipment inspections, monitoring, and alarms, and describes the use of engineering

controls and multiple layers of protection or barriers between phosgene exposure hazards and personnel.

This section includes precautions with regards to piping and valves in phosgene service. The manual

states that users should pay particular attention to

protecting piping from over-pressurization due to liquid phosgene trapped between closed valves;

protecting dry 51 phosgene systems from the intrusion of moisture, which can react with phosgene

and cause severe corrosion and failure; and,

inspecting and testing where stainless steel materials are used to detect the presence of stress

corrosion cracking caused by exposure to chlorides.

The section also states that the use of metallic and non-metallic hoses for permanent or temporary piping

systems may increase the opportunity for phosgene leakage and advises users to give due consideration to

51
Phosgene in the absence of water or moisture, sometimes referred to as anhydrous.

82
E.I. DuPont de Nemours & Co. Final Report September 2011

the design, fabrication, and testing of all components. The manual also notes the potential permeability

issue with PTFE liners, stating that these liners are typically used for phosgene service in well-ventilated

areas; however, it does not specifically describe acceptable methods of ventilation.

4.6.3 National Fire Protection Association (NFPA)


NFPA 55: Compressed Gases and Cryogenic Fluids Code provides fundamental safeguards to users,

producers, distributors, and others who handle compressed gas cylinders and includes general

requirements for storage, occupancy, and emergency response and provisions for specific chemicals or

hazard classes as defined by the NFPA. The current version of the CGA P-1 Standard references NFPA

55 in the Ventilation, Storage, and Site Criteria section for toxic and corrosive gases.

DuPont Belles programs and practices related to the storage and handling of phosgene cylinders does not

align with the provisions set forth in NFPA 55. NFPA 55 defines phosgene as a highly toxic gas because

it contains a lethal concentration (LC50) equal to or less than 200 ppm in air when administered via

inhalation for 1 hour. 52 The LC50 for phosgene is 5 ppm for 1 hour of exposure (CGA P-20, 1995). NFPA

55 includes guidelines for controls in buildings that store compressed gas cylinders, and classifies the

phosgene shed structure as an indoor storage area because the walls comprise more than 25 percent of the

shed perimeter (Figure 12). Indoor storage for highly toxic gases must include a gas cabinet, exhausted

enclosure, or a gas room, according to NFPA 55. Exhausted enclosures, gas cabinets, or gas rooms fully

enclose cylinders and associated process equipment and are equipped with ventilation systems to capture

and treat hazardous vapors. The phosgene shed at Belle, though considered indoor storage by NFPA, does

not contain a ventilation system; instead, DuPont relies on natural ventilation from the outside to decrease

concentrations of phosgene, which allows phosgene vapors to travel downwind, potentially exposing

52
LC50 is the lethal concentration for 50 percent of the exposed population.

83
E.I. DuPont de Nemours & Co. Final Report September 2011

other employees working outside. Without exhausted enclosures, no barriers were present to prevent

phosgene from exposing operators or traveling offsite.

The standard also includes guidance for alarms to warn personnel of potential releases from compressed

gas cylinders and associated equipment. The SLM unit at the Belle plant had alarms for phosgene releases

that were activated manually by the control board operator upon notification from outside personnel or if

a phosgene analyzer activated an alarm at the control board. NFPA 55 guidance states that manual

emergency alarms should be provided in the buildings that enclose cylinders and, when activated, sound

local alarms to alert occupants in the surrounding area. The phosgene shed at Belle contains no alarms

that can be activated locally. Operators suspecting a release are expected to communicate verbally with

the control operator who then sounds an alarm. In the absence of automatic alarm notifications, personnel

in the surrounding area risk exposure, as was the case on the day of the incident.

For gas detection systems, the NFPA states that alarms should activate a local alarm that is both audible

and visual. In the phosgene shed, the SLM building area, and on the Belle Plant fence line, the gas

detection systems activate alarms only in the SLM control room if concentrations exceed the alarm set

points. The gas detectors do not locally sound or visually indicate the detection of a hazardous

concentration to alert surrounding personnel.

4.6.4 Compressed Gas Association (CGA) Standards for the Safe Handling of
Cylinders
The industry association CGA represents manufacturers, distributors, suppliers, and transporters of gases

and cryogenic liquids. It develops and promotes standards and practices for the industrial and medical gas

industry, with input from over 125 member companies. Standards include technical specifications, health

and safety practices, and training and educational materials.

The VanDeMark phosgene bulletin references the current CGA Standard, Safe Handling of Compressed

Gases in Containers (P-1), for the training and proper handling of phosgene cylinders. The 2008 P-1

Standard includes safe practices related to the transportation, identification, and storage of compressed

84
E.I. DuPont de Nemours & Co. Final Report September 2011

gases and specific safe handling and storage rules for chemicals defined by hazard classes. Each

chemical has an assigned hazard class based on its physical properties: flammable, asphyxiant, oxidizer,

toxic, corrosive, or extreme cold. The CGA lists phosgene as a primary toxic and secondary corrosive.

The toxic and corrosive gas section includes requirements for cylinder storage and ventilation, emergency

response, and training. OSHA adopted the 1965 version of the CGA P-1 Standard under the requirements

of the Compressed Gas Standard (29 CFR 1910.101). Under the OSHA Standard, the in-plant handling

and storage of compressed gas cylinders will be in accordance with CGA P-1 (1965).

The current version of the CGA P-1 Standard includes a specific reference to Chapter 7 of NFPA 55 for

the storage and handling of compressed gas cylinders with flammables, but contains only basic

requirements for the storage and handling of corrosives and toxics. In CGA P-1 Section 6.2.6 of

Flammable Gases, the standard includes NFPA 55 requirements such as separation distances, flammable

storage quantities, and fire barriers. However for toxics, the P-1 Standard states, Storage of corrosive

and toxic gases shall be in accordance with local and/or provincial/territorial building and fire prevention

codes. The standard also states that toxics shall be filled and used only in adequately ventilated areas or

preferably outdoors or in exhausted enclosures, but does not contain any specific provisions to achieve

adequately ventilated areas such as the requirements set forth in NFPA Section 7.9.

4.6.5 CGA Standards for PTFE-lined Hoses


On January 29, 2010, CGA published the fourth edition of Standard E-9, Standard for Flexible, PTFE-

lined Pigtails 53 for Compressed Gas Service. Section 1 of E-9 states that the standard applies to hoses

with a diameter of 0.25 inches or smaller and with a maximum allowable working pressure (MAWP) of at

least 3,000 psi, such as those used at DuPont. Section 2 of E-9 states, PTFE-lined pigtails are not

suitable for use with poisonous, toxic, or pyrophoric gases because permeation of gas through the

53
Pigtails are hoses or flexible tubing used to transfer material from a compressed gas cylinder.

85
E.I. DuPont de Nemours & Co. Final Report September 2011

PTFE wall creates a potential hazard. Since phosgene is toxic, this standard rules out using PTFE-lined

hoses for phosgene.

Additionally, Section 5 of Standard E-9 defines how to label hoses: rather than allow tags with adhesive

or heat-shrink wrap, as was the case with the DuPont hoses, it states, The markings shall be made on the

end fitting, collar, separate band, or other permanent location. The hose suppliers practice of affixing

adhesive tape on the hose itself did not align with the requirements in CGA E-9 and enhanced the

corrosion of the metal braid on the PTFE-lined hoses at Belle.

The CGA 2008 P-1 Standard does not specifically reference prior revisions of the E-9 standard. Section

5.9 of P-1 includes general requirements for container connections and states that [p]iping, regulators,

and other apparatus should be kept air tight to prevent leakage... The P-1 Standard does not address

materials of construction or permeability for cylinder discharge hoses in its general or safe handling

requirements by corrosive and toxic hazard class.

4.7 Key Findings


1. An out-of-service phosgene transfer hose failed, exposing a worker to a lethal dose of

phosgene.

2. DuPont did not follow its own standards for the change-out of phosgene transfer hoses.

3. DuPont engineers voiced concern regarding the materials of construction for phosgene hoses

that were not addressed.

4. Liquid phosgene was not evacuated from the riverside hose, as the SOPs indicate, between

transfers to the process from the 1-ton cylinders.

5. A similar hose failure almost occurred a few hours before the exposure of the worker;

however, this near-miss did not prompt an investigation when operators observed the near

failure of the hose on the morning of the fatal release.

86
E.I. DuPont de Nemours & Co. Final Report September 2011

6. The SAP maintenance program was altered so that a work order to change-out the phosgene

transfer hoses was no longer generated automatically (Appendix C).

7. One worker was confirmed to have been exposed to phosgene after the initial exposure while

a second is thought to have been possibly exposed.

8. Emergency responders did not receive timely and detailed information on how to adequately

prepare to respond to the incident.

9. No audible or visual phosgene alarm indication in or around the phosgene shed.

10. The 2009 PHA did not address thermal expansion and corrosion potential for phosgene

transfer hoses.

11. Operators were unaware of the hazards of liquid phosgene thermal expansion (training and

procedures).

12. No plant-wide notification occurred in response to the exposure.

4.8 Root Causes


1. DuPont relied on a maintenance software program to initiate the automatic change-out of

phosgene hoses at the prescribed interval.

2. DuPont did not provide a back-up method to ensure timely change-out of the hoses.

3. A maintenance software program change was not documented or reviewed in accordance

with the MOC process.

4. No person with process knowledge was in place and assigned to convey timely and useful

information to Metro 9-1-1. This responsibility was consigned to the gate guard.

87
E.I. DuPont de Nemours & Co. Final Report September 2011

5. The Belle Plant did not use the construction materials recommended by a corporate expert,

the P3H standard, CGA, or the HTM manual for phosgene hoses, even though the 2006

second-party HTM audit recorded it as an observation.

5.0 Three Incidents in 33 Hours


Because two incidents occurred in a relatively short period, on Saturday, January 23, 2010, after the

oleum release had been secured, the Plant Manager convened a meeting of supervisors and roughly 10

managers and supervisors assigned to the Belle Plant Crisis Committee to discuss and initiate a safety

pause, the intent of which was to evaluate what the managers had seen and take appropriate steps to

ensure safe operation. Approximately 10 managers are part of the Crisis Committee and, after a

debriefing, other supervisors and managers were advised that a safety pause would be conducted. Where

possible, processes would be shut down to allow the discussion, and in those plants that could not be shut

down, employees were expected to participate as best they could.

The Plant Manager assigned the Area Manager for the SLM and F3455 units (who was part of the Belle

Plant Crisis Committee) to contact supervisors and managers and ask that they come to the plant to

participate in planning a plant-wide safety pause. These calls went out at about 11:00 a.m., and

supervisors and managers started arriving at the plant at about noon. At about 2:00 p.m., shortly after the

planning for the safety pause began, the group heard a radio call advising the plant of a medical

emergency. In response to the Plant Managers inquiry, it was learned a worker had been exposed to

phosgene in the SLM unit, making it the third incident in about 33 hours at the facility.

In a striking similarity of events and activities, after two release incidents at the Honeywell Baton Rouge

facility in July 2003, upper management ordered the entire plant to shut down and review all facility

88
E.I. DuPont de Nemours & Co. Final Report September 2011

operations prior to re-start. During this safety stand-down, a third incident occurred where an employee

was exposed to hydrofluoric acid during cleanup of an area in the plant. 54

The objective of both shutdowns was to get the attention of the workforce, acknowledge that the

occurrence of incidents was unacceptable, and recommit to the two companies core values of adhering to

health and safety guidance. One common element was that both companies initiated safety stand-down

activities after the string of incidents started in their respective plants. Another common theme was the

precursor or near-miss events preceding actual incidents. Despite these efforts to address the cause of the

string of incidents at the Belle plant, a fatal incident occurred. At the Belle plant, although investigations

were conducted, near-miss investigations were not immediately responded to on weekends, including the

near catastrophic failure of a separate phosgene transfer hose only hours earlier. Management at all levels

is responsible for fostering an atmosphere of trust and openness and for encouraging the reporting of near-

misses and incidents, as failure to do so could result in non-reporting of near-miss events (CCPS, 1992).

Despite these efforts to address the cause of the string of incidents at the Belle plants, a fatal incident

occurred.

As part of another investigation of the BP Texas City incident in 2005, 55 the CSB examined corporate

oversight of safety management systems and corporate safety culture. As a result of an urgent

recommendation from that same investigation, The Report of the BP U.S. Refineries Independent Safety

Review Panel, The examination of corporate oversight of safety management systems and corporate

safety culture has been conducted as part of another CSB investigation of the BP Texas City incident in

2005 56, and a blue ribbon panel of experts chaired by former Secretary of State James A. Baker was

54
CSB-2003-13-I-LA (Honeywell).
55
CSB 2005-04-I-TX, 2007.
56
CSB 2005-04-I-TX, 2007

89
E.I. DuPont de Nemours & Co. Final Report September 2011

convened. as the result of an urgent recommendation from that same investigation, The Report of the BP

U.S. Refineries Independent Safety Review Panel. While not indicating that the work/safety culture was

irretrievably broken at the Belle facilityand perhaps within the DuPont Corp.the events before and

after the string of incidents in late January 2010 suggest that the safety culture has shifted; is not

operating as it has historically; and could benefit from an extensive examination of all facets of the safety

culture, both within the facility and throughout the corporation.

5.0.1 Additional DuPont Incidents


About 8 months after the series of incidents at the Belle plant triggered this investigation, another

significant release occurred. At about 4:00 p.m. on September 21, 2010, DuPont Belle plant personnel

discovered a methanol leak in a heat exchanger in the methylamines production unit while conducting

regular sampling of the plant's water effluent stream. More than 160,000 pounds of methanol were

estimated to have been released into the Kanawha River over a 24-hour period. This incident occurred

when pressure on the process side of a heat exchanger was increased to a pressure greater than the steam

condensate side of the process. After troubleshooting, operators suspected a leak on the process side of

the heat exchanger and increased steam pressure until samples of the effluent stream confirmed that the

leak had stopped. No employee or community injuries were recorded as a result of this release.

Almost 3 months after the methanol release, on December 3, 2010, at about 2:23 a.m., a fume alert was

sounded in the amines unit at the DuPont Belle, WV, facility announcing a release of monomethylamine

(MMA). The release occurred while two operatorsone senior operator with 34 years of experience at

DuPont and a junior operator with a little over a yearwere sampling MMA from a rail car. One

operator received first- and second-degree chemical burns to his face, while the other inhaled some of the

escaping MMA and received first-degree chemical burns to his face. Both were transported to Charleston

Area Medical Center for 24-hour treatment and observation and released the following day.

90
E.I. DuPont de Nemours & Co. Final Report September 2011

The CSB investigators returned to the Belle facility to assess the MMA release incident. In examining the

equipment, one area of concern was the design of the valves used to isolate the sampling apparatus. As

configured during the sampling operation, only a single block valve isolated the process from the sample

container. This contrasts with industry standards, which suggest the use of double block valves and bleed

vents to assure that the sample piping is clear of hazardous material prior to disconnecting. About 10

pounds of MMA are estimated to have been released during this incident; no employee or community

injuries were recorded as a result of this release.

At DuPonts Yerkes facility in Tonawanda, NY, the CSB assessed a hot work incident that killed a welder

and injured his supervisor on November 9, 2010. This incident was under investigation as this report

went to publication, but preliminary assessments indicate that pre-hot work inspections were less than

adequate, including a failure to check the atmosphere in a tank that normally processes non-flammable

material, but that had inter-connecting piping that could route flammable vinyl fluoride into the tank. The

workers were assigned to repair the tank; however, prior to beginning work, there is no record of DuPont

using a portable gas detector to ensure that the tank being worked on was free of flammable material.

5.1 Management Systems


5.1.1 Knowledge Management
DuPont employees told the CSB investigators that many very knowledgeable Belle plant operations and

maintenance workers had recently retired or are approaching retirement age. From 2005 to the end of

2009, 82 Belle Plant employees retired and 14 resigned. The total number of employees at the Belle plant

has dropped 13 percent (55 people) over the last 5 years. A loss of plant-specific knowledge, or

corporate memory fade, has contributed several incidents in industry (CCPS, 1995), as new hires

cannot replace years of experience; thus, companies must train and supervise new staff until they acquire

job competencies to work safely.

91
E.I. DuPont de Nemours & Co. Final Report September 2011

Experienced maintenance mechanics and technicians have valuable hands-on experience and knowledge

of equipment essential to the safe operation of plant processes. A worker in the Belle maintenance

department told the CSB investigators that the maintenance staff reported to four different maintenance

site leaders over the last 5 years prior to the January 2010 incidents. Other employees expressed concern

that new hires spent too little time learning from veteran employees.

The CSB investigators reviewed and compiled workforce data from DuPont Belle organization

announcements between January 2005 and June 2010, which listed all new hires, transfers, resignations,

and retirements that affected the Belle workforce. Over the 4 years, there were 85 retirements totaling

2,572 years of experience with an average 30 years of service per employee. Among the 85, 20 were from

the maintenance department, contributing to a loss of 713 total years of knowledge and experience (Table

6).

DuPont Belle Workforce 2005 to 2009

Retirements Years Experience at Belle


Maintenance 20 713
Total 85 2,572

New hires Years Experience at Belle


Maintenance 10 0
Total 101 0

Table 6. Sum of Belle plant retirements and new hires from 2005 to 2009 57

In addition to the 85 retirements, there were 14 resignations and 14 transfers to other sites. The Belle plant

hired 101 employees over the 4 years and 8 DuPont employees transferred to Belle from other sites.

Though the overall proportion of new to departing employees has remained consistent, a significant

57
This does not include interns, co-ops, special assignments, or leaves of absence.

92
E.I. DuPont de Nemours & Co. Final Report September 2011

reduction of employees with an average of 30 years of experience working on the Belle site contributes to

a loss of institutional and plant-specific knowledge.

In the case of Belle, a significant population of employees is retiring, with a great deal of process

knowledge that is lost if not properly maintained. This is an issue for industry in general as an entire

generation of baby boomers approaches retirement. In January, 2011, DuPont announced plans to hire

150 employees at Belle over the next few years to compensate for the number of retiring workers.

5.1.2 Hierarchy of Controls


The Hierarchy of Controls is a method generally recognized and used by health and safety professionals

to control workplace hazards. The National Safety Council (NSC) developed the Hierarchy of Controls in

the 1950s and Congress later adopted and enacted it into the Occupational Safety and Health Act of 1970.

The Hierarchy of Controls (Figure 17) demands the use of higher-level engineering and administrative

controls to eliminate hazards. When those operations are not feasible, a PPE program must be

implemented.

Figure 17. Hierarchy of Controls

93
E.I. DuPont de Nemours & Co. Final Report September 2011

In the early 1900s, DuPont recognized that eliminating hazards is preferred beyond education and

protection. However, SOPs for the SLM phosgene cylinder feed system relied primarily on work practices

and PPE to protect operators from the exposure hazards. Other facilities within DuPont and in the

chemical industry have engineering controls in place for similar phosgene cylinder operations.

5.1.2.1 Design and Engineering Controls for Phosgene Cylinders


In 1984, Ciba-Geigy Corp. employees published a technical paper about the safe handling of phosgene in

chemical processing specific to the operation of 1-ton phosgene cylinders (Alspach et al., 1984). Ciba-

Geigy, now part of BASF, had a facility in Toms River, NJ, where two 1-ton cylinders of phosgene fed a

chemical process. Similar to DuPont, the cylinders connected to the process through PTFE-lined hoses

with a stainless steel overbraid induced with nitrogen to drive liquid from the cylinders. At the Ciba-

Geigy plant, a transparent isolation chamber enclosed the cylinder valve connections, and operators

opened and closed valves while standing outside the enclosure, extending their arms though rubber arms

and gloves that were part of enclosure. The enclosure continuously vented to a caustic scrubber and acted

as a barrier between the operator and any potential phosgene vapors near the cylinders.

The phosgene area had phosgene analyzers to continuously monitor and alarm if concentrations exceeded

a defined set point. At high concentrations, flashing lights and audible warnings automatically alerted the

production building, plant guards, and adjacent roadways and buildings. At the Belle facility, phosgene

readings on the analyzers activate alarms in the control room, but DuPont relies on the board operator to

notify personnel in the unit and the rest of the plant. By automating the phosgene analyzer alarm system

to activate notifications plant-wide, Ciba-Geigy eliminated reliance on administrative controls to notify

and protect personnel.

5.1.2.2 Phosgene Handling at the DuPont Mobile, AL Plant


The DuPont Mobile plant in Mobile, AL, uses the same 1-ton phosgene cylinders as Belle for its

agricultural chemicals process. The Mobile process has three cylinders on weigh scales, transferred to the

94
E.I. DuPont de Nemours & Co. Final Report September 2011

process through similar PTFE-lined flexible hoses with a stainless steel overbraid made by a different

manufacturer. The Mobile hoses are 18 inches shorter and have a greater maximum operating temperature

and pressure than those used at Belle. A hose distributor supplies both hoses from the manufacturer to

each site.

The phosgene cylinders and weigh scales at the Mobile plant are housed in the cylinder room, an enclosed

room that vents to an emergency scrubber that pulls a slight negative pressure on the room and scrubs air

before venting to the atmosphere. The scrubber is designed to capture vapors from a release of an entire

cylinder. Operators at the Mobile plant enter the phosgene cylinder area under the same PPE requirements

as Belle for isolating and changing cylinders (hard hat, steel-toed shoes, safety glasses, and phosgene

dosimeter). However, at Mobile, to capture and scrub phosgene vapors in the event of a release, the

operator turns on the emergency scrubber and pump before entering the enclosure.

Like Belle, Mobile has phosgene analyzers located in and around the unit to continuously monitor

concentrations. At Mobile, alarms in the cylinder enclosure activate local audible alarms inside the

enclosure and a flashing light outside to alert employees. If no operators are present in the enclosure when

the alarm activates, the emergency vent scrubber automatically starts. The Belle plant analyzer in the

phosgene shed has no audible alarm to alert personnel in the area; instead, Belle plant procedures require

the board operator to notify personnel of the release and only operators at the phosgene shed can activate

the switch for the warning light.

The emergency scrub system and automated alarms at Mobile are examples of higher-level controls that

protect workers. Mobile has automated alarms where Belle relies on operator action to initiate alarms to

warn personnel of a suspected or actual release. Mobile implemented the scrubber system, an example of

an engineering control, to manage the concentrations of phosgene in the cylinder enclosure in the event of

a release. The Belle plant phosgene shed design allows only for natural ventilation to carry unwashed

95
E.I. DuPont de Nemours & Co. Final Report September 2011

phosgene gases that can potentially harm personnel in or around the shed and possibly enter the

community.

5.1.2.3 Safety in Design Issues


Safety considerations in the equipment design stage eliminate the need for companies to retrofit existing

process equipment or implement administrative or PPE programs to protect workers and the environment.

In addition to the SLM unit, the CSB also identified a lack of safety and health considerations during the

design and construction phases of the F3455 and SAR units. In the F3455 unit, engineers did not design

the control system alarms so that operators could distinguish between a failed battery and activation of a

rupture disc burst sensor, which resulted in nuisance alarms for the rupture disc on the methyl chloride

vent line. Instead of addressing the reliability issues associated with the frequently failing sensor,

management wired the burst sensor to electric power so that low batteries were no longer causing frequent

and false alarms. However, since operators were not retrained to respond to the alarm, they ignored the

alarm during the F3455 unit maintenance activity; consequently, the unit restarted with a failed rupture

disc.

The CSB investigators also noted safety in design issues with the presence of the weep hole on the methyl

chloride vent line upstream of the rupture disc assembly. DuPont engineering standards require that

drainage holes be placed downstream of the relief devices on vent lines to allow for drainage and prevent

liquid from lodging in the discharge side of the rupture disc. However, the location of the weep hole

allowed toxic vapors from the methyl chloride vent line to enter the F3455 building where concentrations

could accumulate to dangerous levels. DuPont could have designed the vent line so that the weep hole

would drain to the exterior of the facility where vapors would dissipate into the atmosphere if a rupture

disc burst.

In the SAR unit, DuPont chose copper steam tracing to prevent the oleum sample line and other process

lines from freezing, even though steam tracing is not the preferred method for oleum service (Dillon,

96
E.I. DuPont de Nemours & Co. Final Report September 2011

1997). Steam tracing can create hot spots that result in an uneven heat distribution in the oleum sample

line, which can accelerate corrosion. Steam tracing in the SAR unit exacerbated the corrosion incident in

the oleum sample line, resulting in a significant release of oleum. Had the SAR unit design engineers

called for electric tracing or replaced the steam tracing, the larger hole in the sample line might not have

formed.

6.0 Regulatory Analysis

6.1 Occupational Safety and Health Administration (OSHA)

6.1.1 Process Safety Management Program


The OSHA PSM Standard (29 CFR 1910.119) requires employers to minimize or prevent the

consequence of catastrophic incidents involving highly hazardous chemicals by applying elements of the

PSM regulation to covered processes. PSM applies to processes using or producing any of the 137 listed

toxic chemicals at or above threshold quantities and processes with flammable liquids or gases onsite in

quantities of 10,000 pounds or more in one location. The PSM Standard applies to the SLM and F3455

units because they contain listed toxic chemicals in excess of the threshold quantities (TQ) specified in

the regulation.

A PHA is one of the 14 elements in the PSM Standard requiring the employer to assess all PSM-covered

processes to identify, evaluate, and control hazards by using one or a combination of several

methodologies listed in the regulation. Furthermore, the standard requires the PHA to address 58

the hazards of the process

58
29 CFR 1910.119(e)(3).

97
E.I. DuPont de Nemours & Co. Final Report September 2011

engineering and administrative controls applicable to the hazards and their interrelationships such

as appropriate application of detection methodologies to provide early warning of releases

consequences of failure of engineering and administrative controls

In the 2009 PHA for the SLM unit, the team did not assess the potential for a phosgene release from a

failed transfer hose due to corrosion or thermal expansion but did consider these issues in process

equipment downstream of the hoses. The team identified that engineering and administrative controls,

such as the PM system and adherence to SOPs, would reduce the likelihood of a phosgene release from

this equipment. However, the team did not assess the consequences caused by the PM system failing to

initiate hose replacements at the proper frequency. In its 2009 PHA for the SLM unit, an audit team did

not address phosgene thermal expansion in the liquid transfer hose; subsequently, in July 2010, OSHA

issued a serious violation to DuPont.

The PSM Standard also requires employers to conduct an MOC for all modifications to process

chemicals, technology, equipment, and procedures; and changes to facilities that affect a PSM-covered

process. The procedures are meant to address the following prior to the change 59:

The technical basis for the proposed change


Impact of change on safety and health
Modifications to operating procedures
Necessary time for the change
Authorization requirements for the proposed change

The MOC also requires that the employees in operations and maintenance affected by the change be

informed of the change and trained prior to the start-up of that process.

59
29 CFR 1910.119(l).

98
E.I. DuPont de Nemours & Co. Final Report September 2011

Investigators found MOC program deficiencies for modifications made to critical equipment on both the

F3455 and SLM units. On the F3455 unit, DuPonts MOC process approved a design for the rupture disc

alarm system that lacked sufficient reliability to minimize the release of flammable methyl chloride. The

unit changed the rupture disc burst sensor on the methyl chloride vent line from battery power to electric

to eliminate battery failure, but failed to assess the reliability of the burst sensors individually. The MOC

process did not evaluate the basis of the modification to verify that it met the intended purpose of

eliminating nuisance alarms caused by battery failure.

DuPont did not perform an MOC review for the changes to the maintenance system that handled the

phosgene hose replacements on the SLM unit. The modification made to the phosgene hose replacement

work orders kept the system from generating a new work order, thus extending phosgene hose use beyond

its planned service life. DuPont stated that knowledge of the change was limited to only a few key SAP

users, but these users lacked training necessary to recognize its impact on hose replacement frequency.

6.1.2 Compressed Gases


The OSHA Standard for Compressed Gases (29 CFR 1910.101) applies to employers that handle, store,

and use compressed gases in cylinders, portable tanks, or tank cars. The standard includes requirements

for cylinder inspections, safety relief devices, and storage and handling of compressed gas cylinders, and

requires employers to handle and store cylinders in accordance with CGA pamphlet P-1 1965, Safe

Handling of Compressed Gases in Containers.

In the 41 years since OSHA adopted the reference standard as part of the Compressed Gas Regulation,

CGA P-1 has been revised 10 times. The current 2008 version is more comprehensive than the OSHA-

adopted 1965 version, which does not list chemicals by hazard class and contains specific safety

information only for flammable and poisonous gases. The current version lists 82 chemicals that fall into

99
E.I. DuPont de Nemours & Co. Final Report September 2011

the primary toxics category, while the 1965 version lists only 13 poisonous gases as defined by the

Interstate Commerce Commission (ICC). 60 The 1965 standard contains the same general information as

the current version, but lacks detailed guidance for facility siting, emergency response, and safety

information specific to various types of chemicals stored in compressed gas cylinders. The 1965 version

includes obsolete and outdated references and lacks references to applicable OSHA regulations, as it was

published prior to the establishment of OSHA. With respect to the issues identified in the phosgene

release investigation, had OSHA adopted the 2008 version of the CGA P-1 Standard, DuPont would have

been accountable for more phosgene storage engineering controls via the incorporation of NFPA 55 and

other consensus standards referenced in the standard.

6.1.3 Inspection History


OSHA is authorized under the Occupational Safety and Health Act of 1970 to inspect workplaces to

ensure that employers are providing a safe and healthy work environment by complying with OSHA

standards. A range of inspection categories establish a system of priorities:

Imminent danger
Catastrophes and fatal accidents
Complaints and referrals
Programmed inspections
Follow-up inspections

60
A regulatory body abolished in 1995, some of whose responsibilities were transferred to the Surface
Transportation Board, an agency within the U.S. Department of Transportation.

100
E.I. DuPont de Nemours & Co. Final Report September 2011

A review of OSHAs inspection history reveals that three planned inspections were conducted at the Belle

facility in 1982, 1984, and 1993, in addition to one unprogrammed-related 61 inspection in 1981. Although

no planned inspections occurred from 1993 through 2010, two inspections, one in 1995 and one in 2004,

were the result of complaints; both were closed.

In a series of post-incident inspections, OSHA cited DuPont for a serious violation of Section 5(a)(1) of

the Occupational Safety and Health Act, alleging that inspections were not conducted for all sections of

oleum piping based on prior leak incidents at the SAR unit. Citations for numerous violations of the

PSM 62 Standard were also issued. OSHA cited DuPont for serious violations, including the company's

failure to properly inspect piping used to transfer phosgene, perform a thorough PHA for its phosgene

operation, and train workers on hazards associated with phosgene. Proposed penalties for all violations

totaled $43,000.The OSHA PSM Standard (29 CFR 1910.119) requires employers to prevent or minimize

the consequences of a catastrophic release of highly hazardous chemicals and of flammable liquids and

gases. Phosgene and methyl chloride are listed chemicals, and the SLM and F3455 units processed more

than the TQ, thus the PSM Standard applied.

6.2 Environmental Protection Agency


The EPA Risk Management Program (RMP) regulation (40 CFR 68), mandated by Section 112(r) of the

Clean Air Act Amendments of 1990, regulates the use of highly hazardous chemicals at fixed facilities.

Its purpose is to prevent accidental offsite releases of listed substances and ensure that a company and the

61
An unprogrammed-related inspection can occur at a multi-employer worksite when an employer is being
inspected because of a complaint, accident, or referral. Any other employer with staff on the worksite is subject to
inspection.
62
PSM is a regulation promulgated by OSHA. A process is any activity or combination of activities including any
use, storage, manufacturing, handling, or the onsite movement of HHCs as defined by OSHA and the EPA.

101
E.I. DuPont de Nemours & Co. Final Report September 2011

community are able to respond effectively in the event of a release. The regulation applies to facilities

using or storing regulated substances exceeding the TQ specified in the EPA regulations.

Each covered process is required to be designated as one of three prevention program levels based on

offsite consequence analyses, incident history, and PSM program applicability. Program 1 is the lowest,

simplest management program; Program 2 is an intermediate management-level program with added

program elements and basic documentation requirements (PSM-covered processes cannot be designated

Program 2); Program 3 is the highest-level management program. Most PSM-covered processes are

Program 3, which requires a rigorous management program with detailed record retention criteria and all

PSM program elements. All PSM program activities and records are directly applicable to Program 3

regulatory activities, and all RMP covered chemicals at the DuPont Belle plant fall into Program 3

requirements (Table 7).

Toxics RMP TQ (lbs)


Anhydrous Ammonia 10,000
Phosgene 500
Sulfur Trioxide 10,000
Formaldehyde 15,000
Oleum 10,000
Methyl Chloroformate 5,000
Flammables
Dimethylamine 10,000
Methylamine 10,000
Methyl Ether 10,000
Ethylamine (70% aqueous) 10,000

Table 7. DuPont Belle RMP-covered chemicals and threshold quantities

Each covered process must undergo a hazard assessment (40 CFR 68, Subpart B) in which the owner is

required to prepare a worst case release scenario and an alternative release scenario for each covered

process. Different analysis criteria apply based on whether the covered chemical is toxic or flammable.

The hazard assessment also requires inclusion of the five year accident history. The results of the

hazard assessment, along with other pertinent information for each covered process, must be submitted to

102
E.I. DuPont de Nemours & Co. Final Report September 2011

the EPA. The RMP (40 CFR 68, Subpart G) is submitted electronically and must be periodically updated.

The DuPont RMP submission for 2010 had no accident history to report.

In November 2003, the EPA Region III Chemical Accident Prevention Program audited the Belle facility

to ensure compliance with the EPA RMP, and covered all RMP elements and emergency response and

site security. The EPA audited the 2-million gallon ammonia storage tank against the RMP requirements

for Program 3 management programs and the RMP documentation DuPont submitted. The EPA audit

report submitted to DuPont in December 2003 contained no deficiencies or recommendations for

improvement. The November 2003 RMP audit is the only one conducted at the Belle Plant prior to the

January 2010 incidents.

6.3 State Hazardous Chemical Release Prevention Program


On January 20, 2011, the CSB Bayer CropScience investigation resulted in a recommendation being

issued to the Kanawha-Charleston Health Department to establish a Hazardous Chemical Release

Prevention Program, whose objective is to enhance the prevention of accidental releases of highly

hazardous chemicals and optimize responses if they occur. In light of its proximity in the Kanawha

Valley, the series of incidents at the DuPont Belle, WV, facility support the plants inclusion in such a

program.

The implementation of the new program would incorporate several key guidelines applicable to chemical

plants operating in the Kanawha County. The Belle facility is one of 13 in the county that report EPA

RMP-covered chemicals assigned as Program level 3 that could fall under the auspices of the new

program. The recommendation to the Kanawha-Charleston Health Department stated:

Specifically, the Bayer report recommends that the Director of the Kanawha-Charleston

Health Department establish a Hazardous Chemical Release Prevention Program to

enhance the prevention of accidental releases of highly hazardous chemicals, and

optimize responses in the event of their occurrence. In establishing the program, study

103
E.I. DuPont de Nemours & Co. Final Report September 2011

and evaluate the possible applicability of the experience of similar programs in the

country, such as those summarized in Section 5.3 of this report. At a minimum:

a. Ensure that the new program:

1. Implements an effective system of independent oversight and other services to


enhance the prevention of accidental releases of highly hazardous chemicals

2. Facilitates the collaboration of multiple stakeholders in achieving common goals


of chemical safety; and,

3. Increases the confidence of the community, the workforce, and the local
authorities in the ability of the facility owners to prevent and respond to
accidental releases of highly hazardous chemicals

b. Define the characteristics of chemical facilities that would be covered by the

new Program, such as the hazards and potential risks of their chemicals and

processes, their quantities, and similar relevant factors;

c. Ensure that covered facilities develop, implement, and submit for review and

approval:

1. Applicable hazard and process information and evaluations.

2. Written safety plans with appropriate descriptions of hazard controls, safety


culture and human factors programs with employee participation, and
consideration of the adoption of inherently safer systems to reduce risks

3. Emergency response plans; and,

4. Performance indicators addressing the prevention of chemical incidents.

d. Ensure that the program has the right to evaluate the documents submitted by
the covered facilities, and to require modifications, as necessary

e. Ensure that the program has right-of-entry to covered facilities, and access to

requisite information to conduct periodic audits of safety systems and

investigations of chemical releases;

104
E.I. DuPont de Nemours & Co. Final Report September 2011

f. Establish a system of fees assessed on covered facilities sufficient to cover the

oversight and related services to be provided to the facilities including necessary

technical and administrative personnel; and,

g. Consistent with applicable law, ensure that the program provides reasonable

public participation with the program staff in review of facility programs and

access to:

1. The materials submitted by covered facilities (e.g., hazard evaluations, safety

plans, emergency response plans);

2. The reviews conducted by program staff and the modifications triggered by

those reviews;

3. Records of audits and incident investigations conducted by the program;

4. Performance indicator reports and data submitted by the facilities, and;

5. Other relevant information concerning the hazards and the control methods

overseen by the program.

Ensure that the program will require a periodic review of the designated agency activities and

issue a periodic public report of its activities and recommended action items. 63

63
CSB-2008-I-WV (Bayer CropScience).

105
E.I. DuPont de Nemours & Co. Final Report September 2011

7.0 Recommendations
The CSB makes recommendations based on the findings and conclusions of its investigations.

Recommendations are made to parties that can effect change to prevent future incidents, which may

include the companies involved; industry organizations responsible for developing good practice

guidelines; regulatory bodies; and/or organizations that have the ability to broadly communicate lessons

learned from the incident, such as trade associations and labor unions.

Phosgene Exposure
The Occupational Safety and Health Administration (OSHA)
2010-06-I-WV-R1

Revise OSHA 29 CFR 1910.101, General Industry Standard for Compressed Gases, to require

facilities that handle toxic and highly toxic materials in compressed gas cylinders to incorporate

provisions that are at least as effective as the 2010 edition of Section 7.9, Toxic and Highly Toxic

Gases, in National Fire Protection Association (NFPA) 55, Compressed Gases and Cryogenic

Fluids Code, including enclosures, ventilation and treatment systems, interlocked fail-safe

shutdown valves, gas detection and alarm systems, piping system components, and similarly

relevant layers of protection.

2010-06-I-WV-R2

Take sustained measures to minimize the exposure of hazards to workers handling highly toxic

gases from cylinders and associated regulators, gages, hoses, and appliances. Ensure that OSHA

managers, compliance officers, equivalent state OSHA plan personnel, and regulated parties

conform, under the Process Safety Management Standard (29 CFR 1910.119) Recognized and

Generally Accepted Good Engineering Practices (RAGAGEP) provisions, to industry practices at

least as effective as the following:

106
E.I. DuPont de Nemours & Co. Final Report September 2011

1. NFPA 55 - Compressed Gases and Cryogenic Fluids Code (2010)

2. CGA P-1 Safe Handling of Compressed Gases in Containers (2008)

3. CGA E-9 Standard for Flexible, PTFE-lined Pigtails for Compressed Gas Service (2010)

4. ASME B31.3 Process Piping (2008)

DuPont Belle Plant


2010-06-I-WV-R3

Improve the existing maintenance management by

Supplementing the computerized system with sufficient redundancy to ensure tracking

and timely scheduling of preventive maintenance for all PSM-critical equipment.

Conducting Management-of-Change (MOC) reviews for all changes to preventive

maintenance orders for all PSM-critical equipment in the computerized maintenance

management system.

2010-06-I-WV-R4

Revise the facility emergency response protocol to require that a responsible and accountable

DuPont employee always be available (all shifts, all days) to provide timely and accurate

information to the Kanawha County Emergency Ambulance Authority (KCEAA) and Metro 9-1-

1 dispatchers.

2010-06-I-WV-R5

Revise the near-miss reporting and investigation policy and implement a program that includes

the following at a minimum:

Ensures employee participation in reporting, investigating, analyzing, and recommending

corrective actions as appropriate for all near-misses and disruptions of normal

operations.

107
E.I. DuPont de Nemours & Co. Final Report September 2011

Develops and encourages use of an anonymous electronic and/or hard copy near-miss

reporting process for all DuPont Belle site employees.

Establishes roles and responsibilities for ownership, management, execution, and

resolution of recommendations from incident or near-miss investigations at the DuPont

Belle facility.

Ensures that the near-miss investigation program requires prompt investigations, as

appropriate, and that results are promptly circulated to well-suited recipients throughout

the DuPont Corp.

Ensures that this program is operational at all times (e.g. nights, weekends, and holiday

shifts).

E.I. DuPont de Nemours and Co., Inc.


2010-06-I-WV-R6

Revise safeguards for phosgene handling at all DuPont facilities by

Requiring that all indoor phosgene production and storage areas, as defined in NFPA 55,

have secondary enclosures, mechanical ventilation systems, emergency phosgene scrubbers,

and automated audible alarms, which are, at a minimum, consistent with the standards of

NFPA 55 for highly toxic gases.

Prohibiting the use of hoses with permeable cores and materials susceptible to chlorides

corrosion for phosgene transfer.

Conducting annual phosgene hazard awareness training for all employees who handle

phosgene, including the hazards associated with thermal expansion of entrapped liquid

phosgene in piping and equipment.

108
E.I. DuPont de Nemours & Co. Final Report September 2011

2010-06-I-WV-R7

Review all DuPont units that produce and handle phosgene that, at a minimum, observe and

document site-specific practices for engineering controls, construction materials, PPE,

procedures, maintenance, emergency response, and release detection and alarms, and use

information from external sources to develop and implement consistent company-wide policies

for the safe production and handling of phosgene.

2010-06-I-WV-R8

For each DuPont facility that uses, but does not manufacture, phosgene onsite

Conduct a risk assessment of manufacturing phosgene onsite against the current

configuration.

Communicate the findings of each assessment to compile recommendations applicable to

all DuPont phosgene delivery systems.

Implement these recommendations.

Compressed Gas Association, Inc.


2010-06-I-WV-R9

Revise CGA P-1, Safe Handling of Compressed Gases in Containers, to include specific

requirements for storing and handling highly toxic compressed gas, including enclosure

ventilation and alarm requirements at least as protective as Section 7.9, Toxic and Highly Toxic

Gases and NFPA 55, Compressed Gases and Cryogenics Fluids Code.

2010-06-I-WV-R10

Revise CGA P-1, Safe Handling of Compressed Gases in Containers, to incorporate by reference

CGA E-9, Standard for Flexible, PTFE-lined Pigtails for Compressed Gas Service.

109
E.I. DuPont de Nemours & Co. Final Report September 2011

American Chemistry Council Phosgene Panel


2010-06-I-WV-R11

Revise the Phosgene Safe Practice Guidelines Manual to

Advise against the use of hoses for phosgene transfer that are constructed of

permeable cores and materials subject to chlorides corrosion.

Include guidance for the immediate reporting and prompt investigation of all

potential (near-miss) phosgene releases.

Methyl Chloride Release


E.I. DuPont de Nemours and Co., Inc.
2010-06-I-WV-R12

Commission an audit in consultation with operations personnel to establish and identify the

conditions that cause nuisance alarms at all DuPont facilities. Establish and implement a

corporate alarm management program as part of the DuPont PSM Program, including measures to

prevent nuisance alarms and other malfunctions in those systems. Include initial and refresher

training as an integral part of this effort.

2010-06-I-WV-R13

Revise the DuPont PSM standard to require confirmation that all safety alarms/interlocks are in

proper working order (e.g., not in an active alarm state) prior to the start-up of all Higher-Hazard

Process facilities.

2010-06-I-WV-R14

Reevaluate and clarify the DuPont corporate MOC policies to ensure that staff can properly

identify and use the distinctions between subtle and full changes and train appropriate personnel

how to properly apply the distinctions on any changes in the policy.

110
E.I. DuPont de Nemours & Co. Final Report September 2011

By the

U.S. Chemical Safety and Hazard Investigation Board

Dr. Rafael Moure-Eraso

Chair

John Bresland

Member

Mark Griffon

Member

William Wark

Member

William Wright

Member

Date of Board Approval

September 20, 2011

111
E.I. DuPont de Nemours & Co. Final Report September 2011

References

Alspach, J.; Bianchi, R.J. Safe Handling of Phosgene in Chemical Processing. Plant/Operations Progress

Jan. 1984; Vol. 3, No. 1.

American Chemistry Council Phosgene Panel. Phosgene Safe Practice Guidelines Manual; American

Chemistry Council: Washington, DC, 2006-2009.

Center for Chemical Process Safety (CCPS). Guidelines for Hazard Evaluation Procedures; American

Institute of Chemical Engineers Center for Chemical Process Safety: New York, NY, 1992.

CCPS. Guidelines for Safe Process Operations and Maintenance; American Institute of Chemical

Engineers Center for Chemical Process Safety: New York, NY, 1995.

CCPS. Investigating Chemical Process Incidents; American Institute of Chemical Engineers Center for

Chemical Process Safety: New York, NY, 1992.

Collins, J.J., et al. Results From the US Industry-Wide Phosgene Surveillance: The Diller Registry;

American College of Occupational and Environmental Medicine, Article: JOM201162, Jan. 2011.

Compressed Gas Association, Inc. (CGA). Safe Handling of Compressed Gases in Containers; P-1 11th

ed., Compressed Gas Association, Inc.: Chantilly, VA.

CGA. Standard for Flexible, PTFE-Lined Pigtails for Compressed Gas Service, E-9; 4th ed.; Compressed

Gas Association, Inc.: Chantilly, VA.

Dillon, C.P. Materials Selector for Hazardous Chemicals. Concentrated Sulfuric Acid and Oleum;

Materials Technology Institute of the Chemical Process Industries, Inc. (MTI): St. Louis, MO,

1997; Vol. 1.

Klein, J.A. Two Centuries of Process Safety at DuPont. Process Safety Progress. American Institute of

Chemical Engineers; Mar. 2009. pp 114-122.

112
E.I. DuPont de Nemours & Co. Final Report September 2011

Lees, F.P. Loss Prevention in the Process Industries; 3rd ed., Elsevier Butterworth-Heinemann:

Burlington, MA, 2005.

Mottle, W.J.; Long J.F.; Morrison, D.E. Industrial Safety is Good Business; John Wiley and Sons: New

York, NY, 1995.

National Fire Protection Association. NFPA 55 Compressed Gases and Cryogenic Fluids Code;

National Fire Protection Association: Quincy, MA, 2010.

PD-USGOV-EPA National Archives and Records Administration, Archival Research Catalog, ARC

Identifier 5519.1 Series: DOCUMERICA: The Environmental Protection Agency's Program to

Photographically Document Subjects of Environmental Concern, compiled 1972 - 1977.

Sittig, M. Handbook of Toxic and Hazardous Chemicals and Carcinogens; 5th ed.; William Andrew

Norwich: New York, NY, 2005.

U.S. Chemical Safety and Hazard Investigation Board (CSB). Investigation Report, Chlorine Release,

July 20, 2003, Contaminated Antimony Pentachloride Exposure, July 29, 2003, Hydrogen

Fluoride Release, August 13, 2003, Honeywell International, Inc., No. 2003-13-I-LA, 2005.

CSB. Investigation Report, Pesticide Chemical Runaway Reaction Pressure Vessel Explosion, Bayer

CropScience, LP, August 28, 2008, No. 2008-08-I-WV, 2011.

113
E.I. DuPont de Nemours & Co. Final Report September 2011

Appendix A: Three Event Logic Tree

114
E.I. DuPont de Nemours & Co. Final Report September 2011

115
E.I. DuPont de Nemours & Co. Final Report September 2011

116
E.I. DuPont de Nemours & Co. Final Report September 2011

117
E.I. DuPont de Nemours & Co. Final Report September 2011

118
E.I. DuPont de Nemours & Co. Final Report September 2011

Appendix B: Historical and Event Timeline

119
E.I. DuPont de Nemours & Co. Final Report September 2011

Appendix C: SAP Program

The DuPont Belle plant uses the SAP R/3 Plant Maintenance module to schedule PM and repair work and

track maintenance costs. Many companies use a Computerized Maintenance Management System

(CMMS) such as SAP Plant Maintenance for this purpose. In particular, companies use the CMMS to

schedule PM to ensure that PSM-critical equipment functions properly. This appendix gives additional

detail on scheduling and completing PM jobs in SAP, and why SAP failed to issue work orders to change

the hoses.

PM keeps plant equipment functioning properly, and to minimize the likelihood of a phosgene hose

corroding and rupturing, DuPont created a PM job in SAP to replace the hoses regularly. The SAP Plant

Maintenance module automatically schedules the job at the frequency DuPont designates.

In the SAP Plant Maintenance module, DuPont created a number for the physical equipment and an

electronic document, or maintenance plan, to store all information about the job. The maintenance plan

is a complex form with many fields. One field, confirmation required, can be clicked on or off. If

this button is off, SAP schedules the first hose change job; waits the specified time indicated in the

interval field, such as 30 days; and then automatically schedules another hose change job. Thus, when

the button is off, by default SAP schedules hose change-outs every 30 days, which, for critical

equipment subject to intermittent operation, is usually the desired option (CCPS, 1995). If this button is

on, SAP requires confirmation that the hoses have been changed. Thus, if the confirmation-required

button is on, SAP schedules hose changes 30 days after the previous change, but opens the possibility

that no one will confirm the completion date in the system, creating a scenario where SAP will not

schedule the hose change at the pre-determined interval.

120
E.I. DuPont de Nemours & Co. Final Report September 2011

Despite the computer-based and administrative controls that SAP and DuPont provided, in late 2006

someone changed the confirmation-required field for the phosgene hoses from "off" to "on"or requiring

confirmation. These administrative controls highlight gaps that contributed to the fatality.

When an SAP user account is created, access is provided according to the work role profile that DuPont

establishes. Only certain users would have had access to change the data in the maintenance plan for the

phosgene hoses.

Programmers are super users who have higher level access than normal users and can write batch

programs to change data, forms, and other SAP computer code that affects multiple pieces of equipment

and multiple plant sites simultaneously. As an administrative control at DuPont, programmers write

computer code in a development box to prevent creating problems in the SAP production box that

normal users see. When the programmer completes the code or downloads it to the sandbox, the process

owners test the change to see that it performs as requested or if it creates a problem. After the process

owners approve the change, the programmer runs the code or downloads it to the production box and

makes the actual change for regular users. These computer controls help ensure the integrity of the

production box for regular users, but were not enough to prevent the Belle Plant fatality.

The CSB discovered evidence relevant to the SAP change:

The SAP work role controls allow programmers, process owners, and specific Belle Plant

employees to access the phosgene hose maintenance plan.

In 2005, the Belle Plant upgraded from SAP R/2 to the newer SAP R/3 partly because SAP R/3

included the new PM module. Converting from the previous CMMS to the SAP PM module was

a large project that involved site personnel who verified the data in spreadsheets before contract

SAP programmers uploaded the data into SAP.

121
E.I. DuPont de Nemours & Co. Final Report September 2011

Based on this evidence, the most likely scenario is that a programmer accidentally changed the

confirmation-required field for the phosgene hoses. The change may have been an unintended effect of a

valid change that DuPont requested or may have been an accidental change that went undetected.

122
E.I. DuPont de Nemours & Co. Final Report September 2011

Appendix D: Phosgene Release Calculations


DuPont initially estimated that 0.7 pounds of phosgene released from the riverside cylinder hose and

associated valving at the time of the rupture. After more detailed calculations, DuPont revised the

estimated release quantity to 2.0 pounds of phosgene. The CSB performed calculations and modeled the

release to verify the phosgene release quantity.

Process Equipment
Figure 18 shows the hose and piping dimensions and the maximum amount of phosgene present in the

piping system associated with the hose failure.

Figure 18. The hose and piping system that supplied phosgene for the release

1) Area of a circle:

or =

2) Volume of a cylinder is equal to the area of the circle, multiplied by the length ( ):

123
E.I. DuPont de Nemours & Co. Final Report September 2011

3) To determine weight, multiply by the density ( ):

= = lbs

The density of phosgene, given that the ambient temperature was 8 C:

Phosgene contained in the 1-inch pipe:

Thus,

Phosgene contained in the 0.5-inch pipe:

Thus,

Phosgene contained in the 0.5-inch valve:

(1/3 the full length of the valve, since it was closed)

Thus,

124
E.I. DuPont de Nemours & Co. Final Report September 2011

Phosgene contained in the 1-inch valve:

(1/3 the full length of the valve, since it was closed)

Thus,

Phosgene contained in the quarter inch, four foot long hose:

Thus,

The sum of phosgene in the system:

Phosgene Dose Calculation


Using this phosgene release quantity (2.067 pounds), the CSB calculated the approximate concentration

of phosgene the fatally injured operator was exposed to. Assuming the operator was 3 feet from the

release and the phosgene instantly vaporized in a spherical fashion from the point of release, the operator

would have received a lethal dose of phosgene in less than one-tenth of a second. This calculation

assumes homogeneous concentration/mixing within the spherical phosgene gas cloud:

125
E.I. DuPont de Nemours & Co. Final Report September 2011

126
E.I. DuPont de Nemours & Co. Final Report September 2011

Vapor Cloud Dispersion Modeling


The CSB used the ALOHA (Area Locations of Hazardous Atmospheres) 5.4.1 program to model the

phosgene release based on the characteristics of the release and atmospheric conditions on the afternoon

of January 23, 2010. The National Oceanic and Atmospheric Administration (NOAA) and the EPA

developed ALOHA to estimate the threat zones associated with hazardous chemical releases from toxic

plumes, fires, and explosions. The user inputs chemical property and weather information and the

program generates a user-defined release scenario that shows the concentration of toxic gases within a

radius of the release source.

The following assumptions were used to model the phosgene release in ALOHA:

Atmospheric and Environmental Conditions:

Atmospheric temperature: 50 F
Wind speed: calm, 1.5 m/s
Wind direction: from the north
Humidity: 66%
Cloud cover: scattered
Surrounding terrain: urban

Release conditions

Chemical: Phosgene
Amount released: 2 pounds
Release type: instantaneous
Height of release: 4 feet

The ALOHA program generated a display of concentration threat zones over a distance downwind from

the source of the release. Using the EPA MARPLOT program, threat zones are displayed over a satellite

map of the area using a GIS interface (Figure 19).

127
E.I. DuPont de Nemours & Co. Final Report September 2011

Figure 19. ALOHA estimate of phosgene concentrations with MARPLOT GIS overlay

The ALOHA program estimated threat zones for three user selected phosgene concentrations:

2 ppm (IDLH) 0.2 miles from release source

0.5 ppm (odor threshold) 0.3 miles from release source

0.2 ppm (ERPG-2 64) 0.4 miles from release source

64
ERPG-2 is the concentration to which all could be exposed for up to 1 hour without experiencing or developing
irreversible or other serious health effects or symptoms that could impair their ability to take protective action
(AIHA, 2008).

128
E.I. DuPont de Nemours & Co. Final Report September 2011

The release estimates from the ALOHA program are based on the weather conditions recorded at the

Charleston Yeager Airport around the time of the January 23, 2010, phosgene release, but may not

accurately represent atmospheric conditions at the plant. The ALOHA program also does not consider the

topography or terrain surrounding the plant. The fence line monitors south and southwest of the phosgene

shed recorded phosgene concentrations between 0 and 0.27 ppm, suggesting phosgene vapor may have

traveled south of the DuPont Belle plant fence line toward the river. The ALOHA threat zone overlay in

Figure 19 displays a model of the worst case release conditions indicating that IDLH concentrations of

phosgene could have been present on the Kanawha River shortly after the release and lower

concentrations could have traveled across the river. The community reported no odors or exposure

symptoms the afternoon of the phosgene release incident.

129
E.I. DuPont de Nemours & Co. Final Report September 2011

Appendix E: Hazard Analysis for Phosgene Use at Belle


(Documents in this appendix are redacted for confidentiality)

List of Acronyms, Abbreviations, and Terminology

dia diameter
flashing instantly vaporizing liquid
IHI Individual Hazard Index
LD50 50% lethal dose
MM million (old notation style)
PHI Process Hazard Index
ppm parts per million

130
E.I. DuPont de Nemours & Co. Final Report September 2011

131
E.I. DuPont de Nemours & Co. Final Report September 2011

132
E.I. DuPont de Nemours & Co. Final Report September 2011

133
E.I. DuPont de Nemours & Co. Final Report September 2011

134
E.I. DuPont de Nemours & Co. Final Report September 2011

135
E.I. DuPont de Nemours & Co. Final Report September 2011

136
E.I. DuPont de Nemours & Co. Final Report September 2011

137
E.I. DuPont de Nemours & Co. Final Report September 2011

138
E.I. DuPont de Nemours & Co. Final Report September 2011

139
E.I. DuPont de Nemours & Co. Final Report September 2011

140
E.I. DuPont de Nemours & Co. Final Report September 2011

141
E.I. DuPont de Nemours & Co. Final Report September 2011

142
E.I. DuPont de Nemours & Co. Final Report September 2011

143
E.I. DuPont de Nemours & Co. Final Report September 2011

144
E.I. DuPont de Nemours & Co. Final Report September 2011

145
E.I. DuPont de Nemours & Co. Final Report September 2011

146
E.I. DuPont de Nemours & Co. Final Report September 2011

147
E.I. DuPont de Nemours & Co. Final Report September 2011

148
E.I. DuPont de Nemours & Co. Final Report September 2011

149
E.I. DuPont de Nemours & Co. Final Report September 2011

150
E.I. DuPont de Nemours & Co. Final Report September 2011

151
E.I. DuPont de Nemours & Co. Final Report September 2011

152
E.I. DuPont de Nemours & Co. Final Report September 2011

153
E.I. DuPont de Nemours & Co. Final Report September 2011

Appendix F: Hard Pipe to Flexible Hose Transition


Correspondence
(Documents in this appendix are redacted for confidentiality)

List of Acronyms, Abbreviations, and Terminology

AgProducts The Agricultural Products Department/Business of DuPont


dry phosgene liquid phosgene without any water, also called "anhydrous" phosgene
engg spec engineering specification
ESD Engineering Services Division of DuPont
SS stainless steel

154
E.I. DuPont de Nemours & Co. Final Report September 2011

155
E.I. DuPont de Nemours & Co. Final Report September 2011

156
E.I. DuPont de Nemours & Co. Final Report September 2011

157
E.I. DuPont de Nemours & Co. Final Report September 2011

158
E.I. DuPont de Nemours & Co. Final Report September 2011

159
E.I. DuPont de Nemours & Co. Final Report September 2011

160
E.I. DuPont de Nemours & Co. Final Report September 2011

161
E.I. DuPont de Nemours & Co. Final Report September 2011

162
E.I. DuPont de Nemours & Co. Final Report September 2011

163
E.I. DuPont de Nemours & Co. Final Report September 2011

164
E.I. DuPont de Nemours & Co. Final Report September 2011

Appendix G: PHA Recommendation Delay Letter


(Documents in this appendix are redacted for confidentiality)

List of Acronyms, Abbreviations, and Terminology

COCL2 Phosgene
FEL Front-end loading
PM Preventive Maintenance
Rec Recommendation
Rx Reactor

165
E.I. DuPont de Nemours & Co. Final Report September 2011

166
E.I. DuPont de Nemours & Co. Final Report September 2011

167
E.I. DuPont de Nemours & Co. Final Report September 2011

168
E.I. DuPont de Nemours & Co. Final Report September 2011

169
E.I. DuPont de Nemours & Co. Final Report September 2011

170
E.I. DuPont de Nemours & Co. Final Report September 2011

171
E.I. DuPont de Nemours & Co. Final Report September 2011

172
E.I. DuPont de Nemours & Co. Final Report September 2011

173